Reproductive factors and risk of breast carcinoma in a study of white and African-American women

Authors

  • Giske Ursin M.D., Ph.D.,

    Corresponding author
    1. Department of Preventive Medicine, Keck School of Medicine, University of Southern California, Los Angeles, California
    2. Department of Nutrition, University of Oslo, Oslo, Norway
    • Department of Preventive Medicine, University of Southern California/Norris Comprehensive Cancer Center, 1441 Eastlake Avenue, Los Angeles, CA 90089
    Search for more papers by this author
    • Fax: (323) 865-0142

  • Leslie Bernstein Ph.D.,

    1. Department of Preventive Medicine, Keck School of Medicine, University of Southern California, Los Angeles, California
    Search for more papers by this author
  • Yaping Wang M.S.,

    1. Department of Preventive Medicine, Keck School of Medicine, University of Southern California, Los Angeles, California
    Search for more papers by this author
  • Sarah J. Lord M.B., B.S.,

    1. Department of Preventive Medicine, Keck School of Medicine, University of Southern California, Los Angeles, California
    Search for more papers by this author
  • Dennis Deapen Dr.P.H.,

    1. Department of Preventive Medicine, Keck School of Medicine, University of Southern California, Los Angeles, California
    Search for more papers by this author
  • Jonathan M. Liff Ph.D.,

    1. Department of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, Georgia
    Search for more papers by this author
  • Sandra A. Norman Ph.D.,

    1. Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania, Philadelphia, Pennsylvania
    2. Department of Biostatistics and Epidemiology, University of Pennsylvania, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Linda K. Weiss Ph.D.,

    1. Cancer Centers Branch, Office of Centers Training and Resources, National Cancer Institute, Bethesda, Maryland
    Search for more papers by this author
  • Janet R. Daling Ph.D.,

    1. Division of Public Health, Fred Hutchinson Cancer Research Center, Seattle, Washington
    2. Department of Epidemiology, School of Public Health and Community Medicine, University of Washington, Seattle, Washington
    Search for more papers by this author
  • Polly A. Marchbanks Ph.D.,

    1. Division of Reproductive Health, Centers for Disease Control and Prevention, Atlanta, Georgia
    Search for more papers by this author
  • Kathleen E. Malone Ph.D.,

    1. Division of Public Health, Fred Hutchinson Cancer Research Center, Seattle, Washington
    2. Department of Epidemiology, School of Public Health and Community Medicine, University of Washington, Seattle, Washington
    Search for more papers by this author
  • Suzanne G. Folger Ph.D.,

    1. Division of Reproductive Health, Centers for Disease Control and Prevention, Atlanta, Georgia
    Search for more papers by this author
  • Jill A. McDonald Ph.D.,

    1. Division of Reproductive Health, Centers for Disease Control and Prevention, Atlanta, Georgia
    Search for more papers by this author
  • Ronald T. Burkman M.D.,

    1. Department of Obstetrics and Gynecology, Baystate Medical Center, Springfield, Massachusetts
    Search for more papers by this author
  • Michael S. Simon M.D., M.P.H.,

    1. Department of Internal Medicine, Karmanos Cancer Institute at Wayne State University, Detroit, Michigan
    Search for more papers by this author
  • Brian L. Strom M.D., M.P.H.,

    1. Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania, Philadelphia, Pennsylvania
    2. Department of Biostatistics and Epidemiology, University of Pennsylvania, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Robert Spirtas Dr.P.H.

    1. Contraception and Reproductive Health Branch, Center for Population Research, National Institute of Child Health and Human Development, Bethesda, Maryland
    Search for more papers by this author
    • Dr. Robert Spirtas was the project officer for the project that generated the results of the current study.


  • This article is a US Government work and, as such, is in the public domain in the United States of America.

  • The views expressed herein do not necessarily reflect the views of the National Institutes of Health or the U.S. Government.

Abstract

BACKGROUND

Few studies have investigated the association between reproductive factors and the risk of breast carcinoma among African-American women. The authors assessed whether the number of full-term pregnancies, age at first full-term pregnancy, and total duration of breastfeeding were associated with similar relative risk estimates in white and African-American women in a large multicenter, population-based case–control study of breast carcinoma.

METHODS

Case patients were 4567 women (2950 white women and 1617 African-American women) ages 35–64 years with newly diagnosed invasive breast carcinoma between 1994 and 1998. Control patients were 4668 women (3012 white women and 1656 African-American women) who were identified by random-digit dialing and were frequency matched to case patients according to study center, race, and age. Adjusted odds ratios and 95% confidence intervals were estimated using unconditional logistic regression.

RESULTS

For white women, the reduction in risk of breast carcinoma per full-term pregnancy was 13% among younger women (ages 35–49 years) and 10% among older women (ages 50–64 years). The corresponding risk reductions for African-American women were 10% and 6%, respectively. Risk decreased significantly with increasing number of full-term pregnancies for both races and both age categories. Duration of lactation was inversely associated with breast carcinoma risk among younger parous white (trend P = 0.0001) and African-American (trend P = 0.01) women. African-American women tended to have more children compared with white women, but parity rates were lower in younger women than in older women in both racial groups. However, breastfeeding was substantially more common in young white women than in young African-American women.

CONCLUSIONS

Overall, parity and lactation had similar effects on breast carcinoma risk in white and African-American women. If younger African-American women now are giving birth to fewer children than in the past, without a substantial increase in breastfeeding, breast carcinoma rates may continue to increase at a more rapid rate among these women compared with white women. Cancer 2004. Published 2004 by the American Cancer Society.

There is substantial epidemiologic evidence that reproductive factors play a major role in the risk of developing breast carcinoma.1, 2 Most studies have been conducted among white women, for whom number of full-term pregnancies, age at first full-term pregnancy, and duration of breastfeeding are strongly predictive of breast carcinoma risk. Until recently, the few existing studies of African-American women have been limited in sample size and design; the largest study included only 600 patients with breast carcinoma, and only 1 study used a population-based approach.3–5 Although these studies suggest that parity and age at first full-term pregnancy have a similar effect on breast carcinoma risk for African-American and white women3–5 and that the effects of lactation may differ between the two racial groups,5 direct assessment of effect modification by race and age has been limited. However, a recent report6 suggests that parity may increase breast carcinoma risk among young African-American women. In the current report, we provide a detailed assessment of the effect of full-term pregnancies and lactation on the risk of breast carcinoma among younger and older white and African-American women in the large, multicenter National Institute of Child Health and Development Women's Contraceptive and Reproductive Experiences (CARE) Study.

MATERIALS AND METHODS

The Women's CARE Study was a population-based, case–control study designed to evaluate risk factors for breast carcinoma among white and African-American women ages 35–64 years in 5 U.S. cities (Atlanta, GA; Seattle, WA; Detroit, MI; Philadelphia, PA; and Los Angeles, CA).

Methods have previously been presented in detail.7 Study protocols were approved by the institutional review board at each participating center.

Case Group

Case patients were U.S.-born English-speaking women newly diagnosed with pathologically confirmed invasive breast carcinoma between July 1994 and April 1998. Case patients were identified by the Surveillance, Epidemiology, and End Results (SEER) cancer registry at each center except for the center located in Philadelphia, where cases were ascertained by field staff who contacted hospitals in the study area. We oversampled younger case patients and African-American case patients to maximize their numbers in the study. In addition, we randomly sampled older white case patients to provide approximately equal numbers of patients in each 5-year age category. Of the 5982 eligible case patients, we interviewed 4575 (76.5%), including 2953 white women and 1622 African-American women.

Control Group

Control patients were U.S.-born English-speaking women who had never been diagnosed with invasive or in situ breast carcinoma. They were identified through random-digit dialing (RDD). A census of potentially eligible women was created for each household contacted. Control participants were selected randomly from the pool of women and were frequency-matched to case patients by study center, race, and 5-year age group. Approximately 82% of residential households called were screened successfully. Of the 5956 eligible women selected as control subjects, we interviewed 4682 (78.6%), including 3021 white women and 1661 African-American women.

Data Collection

All participants were interviewed in person using a structured questionnaire on demographics, reproductive history (including breastfeeding), medical history (including hormone use and body size), family history of malignancy, and other lifestyle factors (such as smoking, alcohol use, and exercise). Information was recorded up to the date of diagnosis (month and year) for case patients and up to the date of initial household contact by RDD for control subjects. Informed consent was obtained from all participants. A subset of nonresponding case patients and control subjects completed a nonresponse questionnaire that also included questions on parity. Among eligible nonresponders, the questionnaire was completed by approximately 15% of case patients and 30% of control subjects. The overall parity findings were consistent with those observed among these nonresponders. Detailed analyses of nonresponders will be presented elsewhere.

Analyses

The reproductive variables we considered included gravidity (ever pregnant vs. never pregnant), parity (no full-term pregnancy vs. full-term pregnancy), number of pregnancies, number of full-term pregnancies, age at first full-term pregnancy, and years since last full-term pregnancy. For these analyses, we define a full-term pregnancy as any pregnancy reported to have lasted > 6 months (26 weeks), which corresponds to the definition used in the Cancer and Steroid Hormone (CASH) Study.8 Age at first full-term pregnancy was defined for each woman as the age at which that pregnancy ended. We also considered the following breastfeeding variables: ever breastfed versus never breastfed, ever breastfed for ≥ 2 weeks versus never breastfed, total duration of breastfeeding (in months), and total duration of breastfeeding without providing supplemental feedings (in months).

We excluded 22 women from the analyses. Six participants (five white control subjects and one African-American case patient) were excluded because it was not known whether they had ever been pregnant. Four participants (three white control subjects and one African-American control patient) who were experiencing their first pregnancy at the reference date were excluded from the analyses of full-term pregnancies. One African-American case patient chose not to answer questions about her last three pregnancies, and two African-American women (one case patient and one control patient) reported unknown gestational length for at least one pregnancy. In addition, we excluded two case patients and one control subject with missing information on breastfeeding, one control subject with missing age at first full-term pregnancy, and three case patients and two control subjects who reported that they had never menstruated. Eleven women who had unknown age at menarche were assigned an age of 12 for that variable. These exclusions left us with 9235 participants (4567 case patients and 4668 control subjects).

Odds ratios (ORs) and 95% confidence intervals (CIs) for the ORs were estimated using unconditional logistic regression methods9 while controlling for possible confounding factors, including age, race, and study center. We selected our potential confounders a priori and included the following variables in all models: age (35–39, 40–44, 45–49, 50–54, 55–59, or 60–64 years), race (white or African American), family history of breast carcinoma (no first-degree family history, first-degree [mother, sister, or daughter] family history, unknown, or adopted), age at menarche (≤ 11, 12, 13, or > 13 years), study center (1 of 5 sites), and education (high school or lower level of education, technical school or some college, or college graduate). Where appropriate, we also adjusted for age at first full-term pregnancy (≤ 19, 20–24, 25–29, or ≥ 30 years) and number of full-term pregnancies (1, 2, 3, 4, or ≥ 5). We estimated ORs per full-term pregnancy for all women combined, as well as separately for parous women. These estimates were almost identical (ORs differed by ≤ 0.01), and only the values for all women combined are presented. All adjustment variables were included as categoric variables in the model. All P values reported are two sided.

Because various risk factors appear to behave differently in younger and older women, possibly a result of the onset of menopause, we analyzed our data separately for women age < 50 years and women age ≥ 50 years. This distinction appeared to be reasonable based on the possible crossover in incidence rates between white and African-American women that occurs between ages 45 and 49 years.10 We also repeated all of our analyses using age 45 years as a cutoff point. To determine whether race was an effect modifier, we tested whether the fit of the baseline multivariate logistic regression model with a single parity variable was improved by fitting separate parity categories for each racial group within each age group. All analyses were performed using EPILOG software (Epicenter Software, Pasadena, CA) or the SAS statistical package (SAS Institute, Cary, NC).

RESULTS

Approximately 65% of study participants were white (2950 case patients and 3012 control subjects), and 35% were African American (1617 case patients and 1656 control subjects). Having had a full-term pregnancy was associated with reduced breast carcinoma ORs for all women (Table 1). Compared with women who had never been pregnant, women who had ever experienced a full-term pregnancy had a 28% reduction in breast carcinoma OR in the younger white women (ages 35–49 years) and a 23% reduction in the older white women (ages 50–64 years). Among African-American women, the corresponding reductions in breast carcinoma ORs were 10% and 14%, respectively. As the number of full-term pregnancies increased, ORs decreased in all age and racial groups (all trend test P values < 0.03). The magnitude of the point estimates for multiparity differed by race. That is, compared with women who had never been pregnant, women who had 5 or more children experienced an OR reduction of approximately 50% in the white patient group and a reduction of approximately 30% in the African-American patient group (data not shown). However, the reduction in OR per pregnancy was similar for white women (13% in the younger group and 10% in the older group) and African-American women (10% in the younger group and 6% in the older group). The tests for effect modification by race did not achieve statistical significance for either age group, and we observed no statistically significant effect modification by age within either racial group.

Table 1. Breast Carcinoma Odds Ratiosa and 95% Confidence Intervals in Association with Reproductive Factors According to Age and Race: Women's Contraceptive and Reproductive Experiences (CARE) Study
CharacteristicAll womenWhite womenAfrican-American women
CaseControlOR95% CICaseControlOR95% CICaseControlOR95% CI
  • OR: odds ratio; CI: confidence interval.

  • a

    Adjusted for the following categoric variables: age, race, education, age at menarche, family history of breast carcinoma, and study site.

Ages 35–49 yrs            
 Ever had full-term pregnancy            
  Never pregnant3512881.00 2782211.00 73671.00 
 Only non-full-term pregnancy1992310.710.55–0.911431620.700.52–0.9356690.740.45–1.21
  Ever had full-term pregnancy167618280.770.64–0.9199611300.720.59–0.886806980.900.63–1.29
   P value for effect modification test by race  0.42         
 No. of full-term pregnancies            
  Never pregnant3512881.00 2782211.00 73671.00 
  14474340.860.70–1.052562770.750.59–0.971911571.130.76–1.69
  27147350.810.67–0.984684840.780.63–0.982462510.910.62–1.33
  33434070.690.56–0.862062490.670.52–0.881371580.790.52–1.19
  41161620.560.42–0.7653830.500.33–0.7463790.710.44–1.14
  5+56900.500.34–0.7313370.300.15–0.5743530.710.41–1.21
   Trend P value  0.0001   0.0001   0.007 
   P value for effect modification test by race  0.24         
 OR per full-term pregnancy  0.880.84–0.93  0.870.81–0.92  0.900.84–0.97
Ages 50–64 yrs            
 Ever had full-term pregnancy            
  Never pregnant2371931.00 1831441.00 54491.00 
 Only non-full-term pregnancy100910.870.62–1.2459480.930.60–1.4641430.870.49–1.57
  Ever had full-term pregnancy200420370.790.65–0.97129113070.770.61–0.987137300.860.57–1.29
   P value for effect modification test by race  0.72         
 No. of full-term pregnancies            
  Never pregnant2371931.00 1831441.00 54491.00 
  13232830.920.71–1.181791570.900.66–1.231441261.040.65–1.65
  26576200.850.68–1.065014490.860.67–1.121561710.810.51–1.27
  34984980.790.62–0.993423580.720.55–0.951561400.990.62–1.56
  42652810.750.58–0.971591850.660.48–0.90106960.940.58–1.52
  5+2613550.570.44–0.741101580.520.37–0.731511970.640.40–1.00
   Trend P value  0.0001   0.0001   0.026 
   P value for effect modification test by race  0.17         
   OR per full-term pregnancy  0.920.89–0.96  0.900.86–0.95  0.940.89–0.99

Using age 45 years as the cutoff point provided results similar to those presented in Table 1. A statistically significant inverse relation, with OR decreasing as number of pregnancies increased, was observed in all four age/racial groups. The OR per full-term pregnancy was 0.88 (95% CI, 0.81–0.95) among white women and 0.88 (95% CI, 0.80–0.97) among African-American women ages 35–44 years and 0.89 (95% CI, 0.85–0.93) among white women and 0.94 (95% CI, 0.90–0.99) among African-American women ages 45–64 years.

In analyses restricted to parous women (Table 2), we observed increased ORs among younger white women whose first pregnancies occurred at age 20 or later, although the test for trend was not statistically significant (P = 0.18). There was no association with age at first full-term pregnancy among young African-American women. Later age at first full-term pregnancy was associated with increasing breast carcinoma OR among older white women (P for trend = 0.0012), but not among African-American women (P for trend = 0.76; P for effect modification by race = 0.044).

Table 2. Breast Carcinoma Odds Ratiosa and 95% Confidence Intervals in Association with Reproductive Factors in Parous Case and Control Patients by Age and Race: Women's Contraceptive and Reproductive Experiences (CARE) Study.
CharacteristicAll WomenWhite womenAfrican-American women
CaseControlOR95% CICaseControlOR95% CICaseControlOR95% CI
  • OR: odds ratio; CI: confidence interval.

  • a

    Adjusted for the following categoric variables: age, race, family history of breast carcinoma, age at menarche, age at first full-term pregnancy, number of full-term pregnancies, education, and study site.

  • b

    Among women with two or more full-term pregnancies. Seven patients for whom the number of years since the last full-term pregnancy was not known were excluded.

Ages 35–49 yrs            
 Age at first full-term pregnancy (yrs)            
  ≤ 194455141.00 1362061.00 3093081.00 
  20–245175611.060.88–1.273083241.411.07–1.852092370.820.64–1.06
  25–293904101.090.88–1.352863171.310.97–1.75104930.970.68–1.38
  ≥ 303243431.060.83–1.362662831.350.98–1.8758600.760.48–1.19
   Trend P value  0.58   0.18   0.28 
    P value for effect modification test by race  0.12         
 Yrs since last full-term pregnancyb            
  ≥ 107688841.00 4245061.00 3443781.00 
  7–91781811.080.84–1.401261231.130.82–1.5752581.040.67–1.61
  3–61822000.960.72–1.271321371.000.70–1.4350630.890.56–1.43
  < 3961270.780.55–1.1157870.670.42–1.0739401.090.63–1.90
   Trend P value  0.25   0.21   0.97 
    P value for effect modification test by race  0.69         
 Ever breastfed            
  No7136421.00 3032821.00 4103601.00 
  Yes96311860.700.60–0.826938480.680.55–0.842703380.700.55–0.88
   P value for effect modification test by race  0.64         
 Ever breastfed for ≥ 2 weeks            
  Never breastfed7136421.00 3032821.00 4103601.00 
  Ever breastfed for < 2 weeks51630.710.48–1.0523300.690.39–1.2328330.710.41–1.21
  Ever breastfed for ≥ 2 weeks91211230.700.60–0.826708180.680.54–0.852423050.690.55–0.88
   P value for effect modification test by race  0.55         
 Total breastfeeding duration (mos)            
  07136421.00 3032821.00 4103601.00 
  < 1931160.700.52–0.9450590.750.49–1.1343570.640.42–0.98
  1–63744160.750.62–0.902552730.740.57–0.951191430.720.54–0.97
  7–233504430.670.54–0.822733420.630.49–0.82771010.660.46–0.93
  24+1462110.630.48–0.831151740.560.40–0.7931370.820.48–1.39
   Trend P value  0.0001   0.0001   0.01 
    P value for effect modification test by race  0.76         
    OR per 12 mos breastfeeding  0.920.86–0.99  0.900.83–0.98  0.960.84–1.09
    P value for effect modification test by race  0.85         
Ages 50–64 yrs            
 Age at first full-term pregnancy (yrs)            
  ≤ 195996761.00 2533031.00 3463731.00 
  20–248508991.030.88–1.216006591.010.81–1.252502401.100.86–1.41
  25–293873081.371.11–1.703112361.421.09–1.8676721.160.78–1.73
  ≥ 301681541.270.85–1.491271091.170.83–1.6641450.930.56–1.53
   Trend P value  0.04   0.0012   0.76 
    P value for effect modification test by race  0.044         
 Ever breastfed            
  No9408911.00 5685371.00 3723541.00 
  Yes106411460.900.79–1.037237700.900.76–1.063413760.870.70–1.09
  P value for effect modification test by race  0.85         
 Ever breastfed ≥ 2 weeks            
  Never breastfed9408911.00 5685371.00 3723541.00 
  Ever breastfed < 2 weeks100871.120.82–1.5258451.220.80–1.8542420.990.62–1.57
  Ever breastfed ≥ 2 weeks96410590.880.77–1.016657250.880.74–1.042993340.860.68–1.08
   P value for effect modification test by race  0.81         
 Total breastfeeding duration (mos)            
  09408911.00 5685371.00 3723541.00 
  < 11841900.970.77–1.211171131.030.77–1.3867770.850.59–1.24
  1–64464860.860.73–1.013203560.840.69–1.021261300.900.67–1.22
  7–233133320.910.76–1.102202270.930.73–1.18931050.860.62–1.20
  ≥ 241211380.910.69–1.2066740.920.63–1.3555640.850.56–1.30
   Trend P value  0.15   0.25   0.29 
    P value for effect modification test by race  0.80         
    OR per 12 mos breastfeeding  0.990.93–1.07  0.980.89–1.07  1.010.91–1.14
    P value for effect modification test by race  0.48         

To avoid confusing the effect of a first birth with that of a later birth, we evaluated the effect of recent full-term pregnancy in women who had given birth to two or more children. Because of the collinearity between age and years since last full-term pregnancy, we restricted these analyses to women ages 35–49 years. We found no evidence that women who had a recent full-term pregnancy were at increased risk of developing breast carcinoma compared with women who had their last full-term pregnancy ≥ 10 years earlier; this finding held true for all patients combined (P for trend = 0.25) and for each racial group considered separately.

Ever having breastfed was associated with a reduced breast carcinoma OR for young parous women in both racial groups (Table 2). The OR among women who had breastfed for ≥ 2 weeks was approximately 30% lower than that of women who never breastfed. In addition, the OR declined with increasing duration of breastfeeding. Although the point estimates were lower for young white women, we found no evidence of a statistical significant effect modification by race in either young (P = 0.76) or older (P = 0.80) women. We observed no dose-response effects among older women, although the point estimates associated with longer breastfeeding were slightly less than 1. The point estimates and trends were essentially identical when we considered total months of breastfeeding without supplements (results not shown).

Within the group of women ages 35–49 years, the reduction in OR associated with breastfeeding was greater among women who had given birth in the past 5 years than among women whose last birth occurred ≥ 5 years ago (Table 3). Among women who had given birth within the past 5 years, those who had breastfed for ≥ 24 months had a 62% reduction in breast carcinoma OR compared with those who had not breastfed at all. Even among women who had given birth > 5 years ago, there was a 30–40% reduction in OR associated with having breastfed for ≥ 24 months, at least among white women.

Table 3. Breast Carcinoma Odds Ratios and 95% Confidence Intervals in Association with Duration of Breastfeeding by Time Since Last Full-Term Pregnancy for Parous Women Ages 35–49 Years
CharacteristicAll women ages 35–49 yrsWhite women ages 35–49 yrsAfrican-American women ages 35–49 yrs
CaseControlOR95% CICaseControlOR95% CICaseControlOR95% CI
  1. OR: odds ratio; CI: confidence interval.

  2. a Adjusted for the following categoric variables: age, race, family history of breast carcinoma, age at menarche, age at first full-term pregnancy, number of full-term pregnancies, education, and study site.

Last full term < 5 yrs ago            
 Total breastfeeding duration (mos)            
  059361.00 28151.00 31211.00 
  < 115190.420.18–0.98890.490.14–1.707100.510.15–1.78
  1–658740.430.24–0.7844480.420.19–0.9514260.370.14–1.02
  7–2377990.380.21–0.6962820.340.16–0.7615170.470.16–1.38
  ≥ 2432530.380.19–0.7727440.380.15–0.95590.310.07–1.36
   Trend P value  0.003   0.02   0.049 
Last full term ≥ 5 yrs ago            
 Total breastfeeding duration (mos)            
  06566071.00 2772671.00 3793401.00 
  < 178980.710.51–0.9842500.750.48–1.1936480.660.41–1.05
  1–63183420.800.65–0.982112250.770.58–1.011071170.810.59–1.12
  7–232753440.710.57–0.882112600.680.51–0.9064840.680.47–1.00
  ≥ 241141580.690.51–0.93881300.580.40–0.8426280.950.53–1.71
   Trend P value  0.0009   0.001   0.08 

To test whether the results would be similar in other age groups, we estimated ORs per full-term pregnancy and ORs per 12 months of breastfeeding for white and African-American women ages 35–39, 40–55, and 56–64 years. We observed no statistically significant effect modification by age in either racial group for these ORs (results not shown).

To further understand the age- and race-specific results regarding parity and lactation, we evaluated these characteristics among control subjects in each of the four age/racial groups (Table 4). Younger white women were the least likely to have been parous. The average number of children among parous women was lowest for younger white women and highest for older African-American women. The percentage difference in average number of children between white women and African-American women was slightly greater in the older age group (17.2%) than in the younger age group (13.6%). Correspondingly, the proportion of women with more than 3 live births and with their first live birth before age 19 years was greater among African-American women than among white women in both age groups. Younger white women were more likely to have breastfed and had breastfed substantially longer than women in other age/racial groups.

Table 4. Frequency Distribution of Parity and Breastfeeding Practices among Control Patients by Age and Race
CharacteristicWhite womenAfrican-American women
35–49 yrs50–64 yrs35–49 yrs50–64 yrs
  1. SE: standard error.

No. of parous women (% of all women)1130 (74.7)1307 (87.2)698 (83.7)730 (88.8)
Mean no. of live births among parous women ± SE2.2 ± 0.032.9 ± 0.042.5 ± 0.053.4 ± 0.08
No. of women (% of all women) with first live birth at age < 19 yrs126 (8.3)170 (11.3)235 (28.2)276 (33.6)
No. of women (% of all women) with > 3 live births120 (7.9)343 (22.9)132 (15.8)293 (35.6)
No. breastfeeding (% of all parous women)848 (75.0)770 (58.9)338 (48.4)376 (51.5)
Mean no. of mos breastfeeding (among women who breastfed) ± SE15.0 ± 0.629.0 ± 0.479.4 ± 0.7610.8 ± 0.73

DISCUSSION

In the current multicenter, population-based case–control study, we observed that breast carcinoma ORs decreased with increasing number of pregnancies and with increasing number of full-term pregnancies in both racial groups. Compared with women who had never had a full-term pregnancy, women who had experienced at least 1 pregnancy had an estimated 20–24% reduction in breast carcinoma risk. This result was statistically significant only among white women. The estimated risk reduction associated with having given birth to many children was slightly greater among white women than among African-American women. However, the OR per pregnancy was similar among younger and older women in both racial groups. Late age at first birth was associated with an increased breast carcinoma OR in white women only. Breastfeeding reduced ORs in both racial groups. The strongest results were observed for younger women.

The strengths of the current study include its size, its population-based design, the recruitment of both white and African-American women, and the use of a standardized questionnaire in a personal interview. The use of control subjects recruited by RDD is a potential limitation. If RDD control subjects were more likely to represent women with many children who stay home and thus are more willing to participate, then we would expect our findings to be biased away from the null hypothesis (i.e., toward the finding of an exaggerated protective effect associated with parity). Some nonresponders completed a short questionnaire, the findings of which suggest that women who participated in the Women's CARE Study may have been more likely than nonparticipants to have four or more children. However, this observation was true to the same extent for case patients and control subjects (results not shown). Data from the National Survey of Family Growth from 199511 suggest that 12.3% of women ages 35–44 years were nulliparous; this is identical to the percentage of control subjects < 50 years in the current study who were nulliparous. In addition, because our results regarding white women are comparable to those of previous studies, control subjects in the current study are likely to represent an unbiased sample with respect to pregnancy history.

There is abundant epidemiologic evidence that parity, multiparity, and early age at first full-term pregnancy decrease breast carcinoma risk.1 The effects of parity and multiparity have tended to be stronger in postmenopausal (or older) women than in premenopausal (or younger) women.1, 12–14 In the current study, older women had more pregnancies than did younger women. Although the effect of ever having experienced a pregnancy was stronger among older women, the relative risk per full-term pregnancy was similar in both age groups. Our findings regarding the protective effects of total parity are similar, although slightly stronger, than the results reported by a large collaborative analysis of > 50,000 cases of breast carcinoma.2 In that analysis, compared with uniparous women who had never breastfed, the relative risk of breast carcinoma associated with having had 5 or more children was 0.73 for women who had never breastfed and 0.64 for women who had ever breastfed.

The protective effect of parity has been observed for white women throughout the world, but only a limited number of studies of African-American women have been conducted. Data from the SEER cancer registries show that the rate of breast carcinoma incidence among young African-American women is higher than the corresponding rate among white women, whereas older white women have a higher breast carcinoma incidence rate than do older African-American women.15 This crossover in incidence curves may be explained by differences in the prevalence of reproductive risk factors,10 if one assumes that these risk factors have similar effects for white women and African-American women. A report from the CASH Study,5 as well as hospital-based case–control studies conducted in the 1970s3 and the early 1980s,4 found that the effects of parity and age at first full-term pregnancy were similar among white women and African-American women. In the current study, 16.9% of white women and 41.1% of African-American control participants had their first full-term pregnancy before the age of 20 years. If early age at first birth exerts most of its protective effect in the postmenopausal years, and if the effect on white and African-American women is similar, then one would expect older African-American women to have lower breast carcinoma rates compared with white women. Nonetheless, in the current study, early birth provided no protection for African-American women, suggesting that the higher proportion of African-American women who had given birth at an early age had no effect on breast carcinoma risk. The effect of having four or five or more full-term pregnancies was, if anything, stronger among white women than among African-American women. Few white women had that many children, however, so this highly protective effect among the few white women was not sufficient to yield a larger decrease in OR per full-term pregnancy among white women compared with African-American women.

In a recent report, Palmer et al.6 suggested that high parity was associated with an increased risk of breast carcinoma among young African-American women. Using data from a prospective study of African-American women (the Black Women's Health Study), they found that compared with having had 1 birth, the relative risk of breast carcinoma associated with having had 4 or more births was 2.4 (95% CI, 1.1–5.1) for women age < 45 years and 0.5 (95% CI, 0.3–0.9) for women ages 45–70 years. Our findings are not consistent with their report, as we found a decreasing risk with increasing number of births in all age and racial groups, regardless of whether we used age 45 years or age 50 years as the cutoff point in our analyses and regardless of whether we used nulliparous or parous women as the reference group. In addition, unlike the Black Women's Health Study,6 we did not find that having recently given birth affected breast carcinoma risk in African-American women. One possible reason for the discrepancies between the current study and that study is selection bias, which may have been present in the current study if young African-American case patients with many children or nulliparous African-American control subjects chose not to participate. However, the findings in the current study are consistent for white women and African-American women. They are also consistent with the results regarding parity and breast carcinoma for white patients in other studies. Differences in study populations represent another possible reason for the discrepancies between these two studies. The Black Women's Health Study enrolled subscribers to a women's magazine. Of the subscribers enrolled, 46.9% had a college degree or further education; in the Women's CARE Study, the corresponding rate among control subjects was 23.8%,16 and among the African-American control subjects included in the current analysis, the corresponding rate was 19.6%. The populations in these two studies are therefore clearly different. It is possible that different screening practices or other cofactors specific to the more educated women in the Black Women's Health Study explained the observed differences in parity results.

In the current study, on average, African-American women had more children than did white women. However, the difference in the mean number of children between racial groups was slightly smaller among younger women compared with older women. If having given birth to a relatively large number of children was one reason for the lower breast carcinoma rates observed among older African-American women in the past, then a decrease in parity among African-American women may result in higher breast carcinoma incidence rates among older African-American women over time.

A transient increase in breast carcinoma risk after a pregnancy has been observed in a number of studies,17, 18 but not in the CASH study,19 and we observed no such effect in the current study as well. A potential problem associated with investigating the effects of recent pregnancy on breast carcinoma occurrence among young women is that control subjects who have experienced a recent pregnancy may be very busy and thus less willing to participate, which would result in an underestimation of recent pregnancies in the control group and an apparent increase in risk in association with such pregnancies. Case–control studies of women of reproductive age should therefore obtain information regarding recent pregnancies from potential control subjects who choose not to participate.

Our results regarding lactation are consistent with growing evidence that lactation plays a role in reducing breast carcinoma risk. In the recently published pooled analysis of data from 47 epidemiologic studies,2 the risk of breast carcinoma associated with ever having breastfed was 0.96 after adjustment for parity and other risk factors. Breast carcinoma risk decreased by 4.3% with every 12 months of having breastfed. We noted a 5% reduction in risk for all women combined per 12 months of having breastfed. As was the case in the current study, the observed reductions in risk in many previous studies have been stronger for younger or premenopausal women.1, 12, 20–23 However, in some studies, this reduction in risk was observed in the postmenopausal years24–26 or was negligible in both age groups.27, 28

Diversity in terms of breastfeeding duration may explain some of the discrepancies in results across studies. In data from the National Survey of Family Growth from 1995, the average duration of lactation for women age < 45 years was 30 weeks among white women and 23 weeks among African-American women.11 Studies conducted in China have found that the strongest protective effects against breast carcinoma risk are associated with long-term lactation, with risk reductions of ≥ 50% among women who breastfed for ≥ 24 months.22, 29

Although some of the protective effects of breastfeeding that we observed in younger women may be attributable to their higher levels of breastfeeding, our results suggest this is only part of the explanation. Although younger white women were more likely to have breastfed and to have breastfed for a longer time compared with older white women, the corresponding differences between age groups were much smaller among African-American women, with the observed risk remaining substantially higher for younger women compared with older African-American women.

Enger et al.21 found that lactation had a stronger protective effect in the first 5 years after a pregnancy. Our findings confirm this observation, although we also observed a protective effect associated with lactation in younger women who had given birth > 5 years ago. Our results confirm that lactation has a rather immediate effect (most of the protection is observed in the first 5 years after birth) but that some of that protective effect persists for a longer time.

The mechanisms associated with the protective effects of parity and lactation have not been completely elucidated. Pregnancies may protect against breast carcinoma risk via differentiation of the terminal duct lobular units,30 via decreases in plasma levels of estrogen31, 32 or progesterone32 or increases in levels of sex hormone–binding globulin,31 or via modification of the changes in estrogen production that occur with age.33 Lactation may protect against breast carcinoma by postponing the resumption of ovulatory menstrual cycles after a pregnancy, by increasing the differentiation of breast tissue,34 by altering estrogen levels in the breast,35 or by allowing excretion of carcinogenic agents from the breast ductal tissue.36

In the current study, on average, young white women breastfed almost twice as long as did their African-American counterparts. We can speculate on the extent to which this difference contributes to maintaining the crossover in incidence rates between white women and African-American women. Perhaps the higher lactation rates observed among young white women contribute to the decreased breast carcinoma incidence rates for these women compared with young African-American women.

In conclusion, our results support the hypothesis that parity and lactation are associated with decreased risk of developing breast carcinoma for both white women and African-American women. Although African-American women, on average, had slightly more children than did white women, the overall effect of multiparity was slightly greater for white women. Nevertheless, the reduction in risk per full-term pregnancy was similar in both racial groups. Within each racial group, the decline in risk per full-term pregnancy was similar for younger women and older women, but the effect of age at first full-term pregnancy on breast carcinoma risk was only observed among white women. The protective effect of lactation, which was limited to younger women and was stronger during the first 5 years after a pregnancy, was similar in both racial groups. Nonetheless, African-American women were substantially less likely to have breastfed and did so for shorter durations than did their white counterparts.

Our findings suggest only slight differences between white women and African-American women in terms of the effects of reproductive factors on breast carcinoma risk. Even so, the slight differences that we observed may provide important clues regarding the mechanisms underlying the protective effects of pregnancy and lactation against breast carcinoma. If breastfeeding rates remain low and the parity rate continues to decrease over time among young African-American women, a more rapid increase in breast carcinoma incidence rates in this group compared with white women could result. Breastfeeding, and doing so for a longer duration, could reduce the risk of breast carcinoma and should be encouraged, especially among young African-American women.

Ancillary