Reproductive risk factors for breast cancer and association with novel breast density measurements among Hispanic, Black, and White women

Abstract

Purpose:

There are differences in the distributions of breast cancer incidence and risk factors by race and ethnicity. Given the strong association between breast density and breast cancer, it is of interest describe racial and ethnic variation in the determinants of breast density.

Methods:

We characterized racial and ethnic variation in reproductive history and several measures of breast density for Hispanic (n=286), non-Hispanic Black (n=255), and non-Hispanic White (n=1,694) women imaged at a single hospital. We quantified associations between reproductive factors and percent volumetric density (PVD), dense volume (DV), non-dense volume (NDV), and a novel measure of pixel intensity variation (V) using multivariable-adjusted linear regression, and tested for statistical heterogeneity by race and ethnicity.

Results:

Reproductive factors most strongly associated with breast density were age at menarche, parity, and oral contraceptive use. Variation by race and ethnicity was most evident for the associations between reproductive factors and NDV (minimum p-heterogeneity:0.008) and V (minimum p-heterogeneity:0.004) and least evident for PVD (minimum p-heterogeneity:0.042) and DV (minimum p-heterogeneity:0.041).

Conclusion:

Reproductive choices, particularly those related to childbearing and oral contraceptive use, may contribute to racial and ethnic variation in breast density.

INTRODUCTION

In the United States, breast cancer incidence is highest among non-Hispanic White and non-Hispanic Black populations and lowest among Hispanic and Asian populations [1, 2]. Along with societal factors (e.g., social support, toxin exposure, access to high quality care) and cellular factors (e.g., differences in estrogen metabolism and the frequency of genetic risk variants), population-level differences in the distributions of breast cancer risk factors are hypothesized to contribute to racial and ethnic disparities in breast cancer incidence [3–9]. These differences are apparent in the distributions of biologic, reproductive, and lifestyle risk factors [3–5, 7–9]. Investigating how breast cancer risk factors are associated with intermediate markers of breast cancer risk could advance understanding of possible intervention opportunities for reducing racial and ethnic disparities in breast cancer.

Breast density (BD) can be measured from a clinical mammogram where stromal and epithelial tissues are radiodense and appear light, while adipose tissue is radiotransluscent and appears dark. BD is the amount of bright (dense) tissue in an image, noting there are various ways of quantifying dense breast tissue. BD is one of the strongest predictors of breast cancer risk; women with area-based measures of BD >75% are estimated to have four to six times the risk of breast cancer compared to those with BD <5% [10]. Given the strong association between BD and breast cancer, it is of significant public health interest to identify the major determinants of BD and consider how they are distributed across the population.

Prior research has shown that BD is influenced by genetic and non-genetic factors. The overall heritability of BD is estimated to fall between 53% and 67% [11], while other strong determinants of BD include age [12–14], menopausal status [12–14], and BMI [15–19]. Associations between reproductive factors and BD are still under investigation. For example, studies of older age at menarche have reported either a weak, positive association with higher BD [20–24] or no association [15, 16, 25–29]. Results for nulliparity also have been mixed with most studies reporting positive associations with BD [15, 16, 18, 20, 25, 30], but others reporting no association [22, 26, 28]. Among parous women, having fewer children has often [14, 16, 20, 21, 27, 28, 30], but not always [22, 26, 29] been associated with higher BD; older age at first birth has been associated with higher BD in some [15, 18, 21, 22, 25, 30], but not all [16, 26, 27, 29], studies, and the association between breastfeeding and BD has been reported as either weakly positive [26, 30], or null [27, 29]. Oral contraceptive (OC) use [21, 26, 28, 29] and menopausal hormone therapy (HT) use [15, 16, 18, 20, 21, 26, 29, 31–37] have often, but not always, been positively associated with BD.

Innovative developments in BD research that may help to clarify associations between reproductive factors and BD include the incorporation of BD measurements that capture vendor-neutral volumetric characteristics (e.g., automated measures of absolute dense volume [DV], absolute non-dense volume [NDV], and percent volumetric density [PVD]), and measurement of variation (V), which is a measure of pixel intensity variation that is associated with breast cancer risk independent of DV and percent density [38–43]. In the present study, we used full field digital mammography (FFDM) data from a multiracial and multiethnic clinical cohort to characterize DV, NDV, PVD, and V by race and ethnicity, and we investigated whether reproductive risk factors for breast cancer are associated with BD measures.

METHODS

Study population

We included women enrolled in the Boston Mammography Cohort Study (BMCS). Details of this population have been described previously [44]. Briefly, we enrolled 2,821 women with a least one screening mammogram scheduled at Brigham and Women’s Hospital (BWH) between 2006 and 2014. At enrollment, participants completed a baseline questionnaire covering lifestyle, demographic, and reproductive characteristics. They also provided consent to access their BWH mammograms and medical records. For inclusion in BD analyses, women had to have at least one unprocessed digital mammogram stored at BWH (n=2,696).

Exposure and Covariates

Reproductive factors queried on the baseline questionnaire included: age at menarche, parity, age at first birth, lifetime duration of breastfeeding, OC use, and postmenopausal HT use. Covariate data, also collected from the baseline questionnaire, included: age, self-reported race and ethnicity, menopausal status, BMI at age 18, current BMI, history of smoking, alcohol use, personal history of benign breast disease (BBD), and family history of breast cancer in a mother, sister, or grandmother.

Outcome

We collected mammographic images from each participant’s electronic medical record, and used image file header information to determine the date and type of each visit (i.e., screening or diagnostic). Participants without an unprocessed screening mammogram within three months of enrollment were excluded from analyses. For the remaining participants, the unprocessed screening mammogram obtained closest to enrollment was used as the study image (i.e., the baseline mammogram). All mammograms were acquired with General Electric (GE) Senographe 2000D and Hologic Selenia FFDM units at BWH. BD measures were averaged over all available craniocaudal (CC) and mediolateral (MLO) mammography images from both breasts to obtain single, study-level measures.

The three volumetric BD measures (NDV, DV, and PVD—calculated as DV/(NDV+DV)) were produced by fully automated Volpara software (version 1.5.0, Volpara Health Technologies Ltd, Wellington, New Zealand). Because volumetric BD measures were developed more recently and we wanted to be able to compare findings with the existing literature, we also obtained area-based BD measures (PD, DA, and NDA) from each image using LIBRA software (version 1.0.4; University of Pennsylvania, Philadelphia, PA https://www.cbica.upenn.edu/sbia/software/LIBRA/) [45, 46].

V captures variation in the grayscale intensity of a mammogram. It is calculated as the standard deviation of pixels within an eroded breast area; the measurement is constrained to the area corresponding to where the breast was in contact with the compression paddle during image acquisition. For this study, V was estimated in the Heine lab, as described previously [39–42]. Specific to this study, General Electric (GE) Senographe 2000D and Hologic Selenia FFDM units have different pitch and data scale representations. A series of processing steps were taken to make V compatible. First, images from Hologic units were filtered, and pixel spacing was increased to 100μm to match the pitch of the GE images. This filtering was based on the Nyquist criteria, and cubic convolution was used for the pixel spacing interpolation. V was calculated from each mammogram from both types of units forming two sets of measurements. Secondly, to account for pixel scale differences, each set of V measurements was mapped (i.e., two mappings) to a zero mean – unit variance normal distribution.

Statistical analysis

This cross-sectional study captured questionnaire and mammography data from 2,696 women. Of these women, we excluded 206 due to personal cancer history, 74 due to inability to calculate V from their raw mammography image(s), and 21 because they did not have a baseline mammogram. We further excluded 12 women who were missing data on race and ethnicity. This left 2,383 women in the study population (Table 1), 2,235 of whom identified as part of racial and ethnic groups of adequate size to be included in statistical analyses (i.e., Hispanic, non-Hispanic Black, non-Hispanic White).

Table 1.

Age-adjusted characteristics of BMCS participants at baseline by menopausal status, race, and ethnicity

	Premenopausal				Postmenopausal
	Hispanic (n=177)	Non-Hispanic Black (n=112)	Non-Hispanic White (n=784)	Less common groupsa (n=85)	Hispanic (n=109)	Non-Hispanic Black (n=143)	Non-Hispanic White (n=910)	Less common groupsa (n=63)
Percent volumetric density (PVD)	9.9 (6.2)	8.7 (5.6)	14 (7.4)	12.7 (7.7)	6.9 (4.4)	6.7 (5.6)	8.3 (5.6)	7.8 (5.4)
Absolute dense volume (DV, cm3)	73 (40.7)	76.5 (33.1)	74 (41.5)	60.6 (35)	54.8 (26.4)	60.4 (29.3)	52.9 (29.7)	51.7 (25.8)
Absolute non-dense volume (NDV, cm3)	845.5 (493.1)	1113.6 (696.2)	612.8 (465.9)	637.7 (560.3)	934.9 (464.2)	1150.3 (627.7)	766.9 (476.1)	865.3 (588.8)
V metric	0.1 (0.9)	−0.2 (0.8)	0.4 (0.9)	0.3 (0.9)	−0.2 (1.1)	−0.6 (1)	−0.2 (0.9)	−0.3 (0.9)
Age at baseline (years)*	45.3 (5.2)	45.3 (6.5)	45.6 (5)	45.9 (5.7)	57.6 (9.1)	59.6 (7.9)	60.5 (8.3)	60.6 (7.2)
Age at menarche (years)	12.6 (1.6)	12.8 (1.6)	12.9 (1.4)	12.6 (1.4)	12.6 (1.6)	13 (1.9)	12.7 (1.4)	12.8 (1.8)
Parous, %	89.9	84.6	78.8	86.3	92.6	94.1	78.1	90.0
Number of birthsb	2.6 (1.1)	2.6 (1.6)	2.2 (0.9)	2.1 (1)	2.7 (1.1)	2.9 (1.7)	2.3 (1.1)	2.7 (1.4)
Age at first birth (years)ab	25.5 (6.4)	25 (6.7)	32 (5)	31.1 (6.6)	23.7 (4.9)	21.8 (5.4)	28.1 (5.8)	26.1 (7)
Years from menarche to first birth	13.1 (6.5)	12.1 (7.1)	19.1 (5.1)	18.4 (6.5)	11.1 (5)	8.9 (5.9)	15.4 (6)	13.5 (7.2)
Breastfeedingb
− Parous, did not breastfeed, %	20.1	15.1	13.3	9.9	26.1	38.7	30.6	33.1
− Breastfed <6 months, %	41.0	38.0	20.9	37.7	45.0	26.8	24.5	26.7
− Breastfed 6+ months, %	39.0	46.9	65.8	52.4	28.9	34.5	44.8	40.1
Duration of oral contraceptive use
− Never used OCs, %	22.8	21.7	14.4	32.5	32.8	36.0	29.1	32.2
− >0–5 Years, %	52.5	44.8	35.7	38.4	50.5	40.6	38.7	32.7
− 6–10 Years, %	15.4	11.8	22.8	19.3	10.4	11.4	18.2	7.9
− ≥11 Years, %	9.3	21.7	27.2	9.9	6.2	11.9	14.0	27.2
Hormone therapy usec
− Never, %	-	-	-	-	79.9	72.2	57.3	68.6
− Past, %	-	-	-	-	13.3	24.4	32.4	28.2
− Current, %	-	-	-	-	6.9	3.4	10.3	3.2
BMI (kg/m2) at age 18	21.5 (4.9)	21.9 (4.4)	21 (2.9)	21.1 (3.4)	20.5 (3.4)	21.5 (4.6)	21.2 (3.2)	20.8 (4.5)
BMI (kg/m2) at baseline	28.8 (6.3)	31.4 (7.8)	25.1 (5.2)	25.9 (5.9)	28.9 (6.1)	30.9 (7.5)	26 (5.3)	28.3 (7.6)
Alcohol use >1 drink per week, %	16.7	13.6	51.3	10.6	13.3	19.6	48.7	25.6
Smoking
− Never, %	75.6	75.3	67.8	78.0	70.3	54.0	52.7	53.1
− Past, %	17.7	13.5	27.3	12.4	20.7	28.2	42.4	36.9
− Current, %	6.7	11.1	4.9	9.6	9.0	17.9	4.9	10.0
Benign breast disease, %	8.8	8.6	11.0	10.0	15.3	6.8	20.0	10.4
Family history of breast cancerd, %	29.1	25.4	39.0	15.6	19.0	28.8	36.9	20.4

Values are means(SD) for continuous variables and percentages for categorical variables. Values are standardized to the age distribution of the study population unless otherwise indicated (*).

Values of polytomous variables may not sum to 100% due to rounding.

^aIncludes women identifying as Asian, American Indian Alaska Native, Native Hawaiian or Pacific Islander, multiracial, and “other”.

^bNumber of births, age at first birth, and breastfeeding duration among parous women.

^cHormone Therapy use among postmenopausal women.

^dFamily history of breast cancer in a mother, sister, or grandmother

We used multivariable-adjusted linear regression to estimate the associations between reproductive factors and BD features. For each analysis, we restricted the data set to include only those women with data on the risk factor of interest. Since BD differs between pre- and postmenopausal women, and because we were interested in effect modification by race and ethnicity, all analyses were carried out within levels of race, ethnicity, and menopausal status. In each model, we adjusted for age at enrollment (continuous), current BMI (continuous, with a quadratic term to account for non-linearity), BMI at age 18 (continuous), history of smoking (never, past, current), alcohol use (0–1 vs. >1 drink per week), personal history of benign breast disease (ever, never), family history of breast cancer (yes, no), and mutually adjusted for all other reproductive exposures. Given the non-normal distributions of PVD, DV, NDV, PD, DA and NDA, we log-transformed these outcome variables before running models. We used the resulting regression parameters (βs) to calculate the percent difference in outcome per unit change in exposure variable (i) via the following formula: $100*(e^{{\beta }_i}-1)$. As noted in the methods, pre-processing ensures that V is normally distributed, so βs for V were interpreted directly from model output (i.e., as the change in the standard deviation of pixel intensity per unit difference in reproductive factor). All associations are presented with their corresponding 95% confidence intervals (CIs).

We considered how estimates of association varied by race and ethnicity by visually comparing estimates and their corresponding 95% CIs. For each reproductive factor we also conducted two F-tests, one in premenopausal women and one in postmenopausal women, to formally test the null hypothesis of a homogenous association between the reproductive factor and BD measure among Hispanic, non-Hispanic Black, and non-Hispanic White women. The F-tests compared models with interaction terms between the reproductive factor and race and ethnicity to corresponding models without any interaction terms.

Analyses were conducted using SAS version 9.4.1. Tests of statistical significance were 2-sided with α=0.05.

RESULTS

Characteristics of BMCS participants (Table 1) and distributions of BD measures (Figure 1) are presented by menopausal status and by categories of self-reported race and ethnicity. Distributions of reproductive factors differed for pre- versus postmenopausal women, potentially due to variation in nutrition, contraceptive methods, and breastfeeding recommendations across birth cohorts. BD measures also varied by menopausal status; DV was consistently lower and NDV consistently higher among postmenopausal women compared to premenopausal women.

Figure 1.

This figure depicts distributions of breast density measures by menopausal status and self-reported race and ethnicity. The (+) symbol indicates the mean. The middle line of each box shows the median. The upper and lower bounds of each box are the 75^th and 25^th percentiles, respectively. And the error bars are indicative of the upper fence (75^th percentile + 1.5*interquartile range) and lower fence (25^th percentile −1.5*interquartile range).

While age at menarche was similar across Hispanic, non-Hispanic Black, and non-Hispanic White women, the distributions of all other reproductive factors varied across these racial and ethnic groups (Table 1). Parity was greater among Hispanic and Non-Hispanic Black women when compared with non-Hispanic White women, and age at first birth was lower among Hispanic and non-Hispanic Black women compared to non-Hispanic White women. Longer durations of breastfeeding were more common among non-Hispanic White women and less common among Hispanic women. OC use and HT use were less common among Hispanic women and more common among non-Hispanic White women. Distributions of health and lifestyle factors also varied by self-reported race and ethnicity.

For the volumetric measures, PVD was lowest among non-Hispanic Black women and highest among non-Hispanic White women (Figure 1). Since DV was relatively similar across racial and ethnic groups, variation in PVD seemed to be driven by greater NDV among non-Hispanic Black women and lower NDV among non-Hispanic White women. V was lowest among non-Hispanic Black women.

Since PVD was calculated from DV and NDV, the reproductive factors associated with PVD (Table 2; Supplementary Figure 2) reflected those associated with DV, NDV, or both. Reproductive factors associated with DV in the full study population included nulliparity and the number of births among parous women (Table 3). We observed a positive association between nulliparity and DV among premenopausal non-Hispanic Black women (38.0% [1.3% to 88.0%]), but not among premenopausal Hispanic women (15.5% [−15.0% to 56.9%]) or premenopausal non-Hispanic White women (−3.3% [−12.7% to 7.1%]; p-heterogeneity=0.041). We also observed a positive association between nulliparity and DV among postmenopausal women (16.4% [7.9% to 25.6%] higher DV) with estimates ranging from 16.2% (−42.2% to 21.6%) lower among postmenopausal Hispanic women to 28.5% (−6.9% to 77.5%) higher among postmenopausal non-Hispanic Black women (p-heterogeneity=0.20). Among parous women, DV further decreased with number of births among premenopausal women (e.g. 14.0% [−23.1% to −3.7%] lower DV among those with three or more births compared to those with just one birth), but not postmenopausal women. Finally, while no association was detected between breastfeeding and DV in the full study population, we observed a strong inverse association between breastfeeding and DV among premenopausal, but not postmenopausal, Hispanic women.

Table 2.

Multivariable-adjusted^a percent difference in percent volumetric density (PVD) comparing categories of reproductive factors

	Premenopausal									Postmenopausal
	All premenopausal			Hispanic		NH Black		NH White	p-het	All postmenopausal			Hispanic		NH Black		NH White	p-het
	n	% difference (95% CI)	n	% difference (95% CI)	n	% difference (95% CI)	n	% difference (95% CI)		n	% difference (95% CI)	n	% difference (95% CI)	n	% difference (95% CI)	n	% difference (95% CI)
Age at menarcheb
Age 10 or earlier	46	0.0 (−15.7, 18.5)	15	6.0 (−24.9, 49.5)	6	15.2 (−34.9, 104.1)	25	−7.5 (−26.2, 15.9)		86	−16.0 (−26.3, −4.1)	8	−11.9 (−46.2, 44.4)	19	−12.7 (−37.0, 21.0)	59	−19.2 (−31.2, −5.0)
Age 11–12	408	-Reference-	64	-Reference-	45	-Reference-	299	-Reference-		439	-Reference-	48	-Reference-	32	-Reference-	359	-Reference-
Age 13 or later	609	7.7 (0.7, 15.2)	94	2.2 (−14.4, 22.1)	59	18.5 (−5.6, 48.6)	456	8.1 (0.1, 16.8)	0.44	621	0.4 (−6.0, 7.3)	51	−0.2 (−20.9, 25.7)	86	−2.6 (−22.6, 22.4)	484	0.7 (−6.5, 8.5)	0.95
Parityc
Parous	765	-Reference-	137	-Reference-	77	-Reference-	551	-Reference-		829	-Reference-	78	-Reference-	111	-Reference-	640	-Reference-
Nulliparous	176	−9.1 (−16.3, −1.2)	15	4.0 (−23.2, 40.9)	13	−8.7 (−34.7, 27.7)	148	−8.5 (−16.4, 0.0)	0.66	193	12.3 (3.9, 21.4)	8	−1.4 (−34.3, 48.0)	7	11.5 (−26.2, 68.4)	178	13.7 (4.8, 23.4)	0.56
Number of birthsd
1	165	-Reference-	19	-Reference-	17	-Reference-	129	-Reference-		173	-Reference-	11	-Reference-	22	-Reference-	140	-Reference-
2	332	−6.2 (−14.6, 2.9)	54	15.7 (−14.3, 56.2)	27	−3.3 (−32.4, 38.3)	251	−8.2 (−17.2, 1.7)		330	5.4 (−3.9, 15.5)	20	41.0 (−10.1, 121.2)	32	4.0 (−24.2, 42.6)	278	4.2 (−5.8, 15.3)
≥3	268	−6.1 (−15.4, 4.1)	64	12.8 (−18.6, 56.2)	33	−6.2 (−36.9, 39.2)	171	−7.5 (−17.6, 3.8)	0.95	326	2.2 (−7.5, 13.0)	47	27.4 (−16.5, 94.4)	57	−2.9 (−29.2, 33.1)	222	3.5 (−7.7, 16.1)	0.94
Age at first birthd
<25	147	-Reference-	66	-Reference-	46	-Reference-	35	-Reference-		335	-Reference-	49	-Reference-	83	-Reference-	203	-Reference-
25–29	161	−4.6 (−16.9, 9.4)	31	19.7 (−9.0, 57.5)	13	−7.8 (−38.6, 38.4)	117	−11.5 (−27.9, 8.5)		233	8.4 (−2.3, 20.3)	17	7.0 (−26.2, 55.2)	17	5.9 (−26.4, 52.2)	199	8.0 (−4.1, 21.7)
≥30	451	−9.0 (−23.8, 8.7)	36	16.8 (−23.9, 79.2)	18	−25.6 (−59.9, 38.3)	397	−14.2 (−32.5, 8.9)	0.41	256	13.8 (−2.4, 32.6)	12	49.4 (−16.8, 168.3)	10	33.2 (−24.9, 136.3)	234	11.2 (−6.4, 32.2)	0.87
Time from menarche to first birthd
<15 years	227	-Reference-	84	-Reference-	55	-Reference-	88	-Reference-		455	-Reference-	59	-Reference-	86	-Reference-	310	-Reference-
≥15 years	528	13.6 (−2.1, 31.8)	47	2.4 (−31.1, 52.2)	21	64.4 (−8.7, 196.0)	460	15.6 (−2.7, 37.2)	0.25	360	−1.6 (−13.2, 11.5)	18	−5.8 (−45.0, 61.2)	21	−5.1 (−44.6, 62.6)	321	−0.9 (−13.5, 13.6)	0.96
Breastfeedingd
Parous, did not breastfeed	111	-Reference-	26	-Reference-	14	-Reference-	71	-Reference-		242	-Reference-	15	-Reference-	36	-Reference-	191	-Reference-
>0–6 months	196	6.8 (−4.9,20.0)	55	−26.3 (−43.4, −3.9)	29	13.2 (−24.4, 69.5)	112	13.1 (−2.1, 30.7)		210	−0.2 (−9.0, 9.6)	34	3.0 (−30.6, 52.9)	25	−7.7 (−34.3, 29.6)	151	3.4 (−7.0, 15.1)
>6 months	437	8.2 (−2.9, 20.4)	52	−24.3 (−42.7, −0.1)	32	17.2 (−24.0, 80.7)	353	15.2 (1.6, 30.6)	0.09	334	3.9 (−5.0, 13.6)	22	4.6 (−28.9, 53.8)	36	−0.4 (−27.8, 37.5)	276	5.6 (−4.4, 16.7)	0.53
OC useb
Never	170	-Reference-	36	-Reference-	23	-Reference-	111	-Reference-		335	-Reference-	31	-Reference-	44	-Reference-	260	-Reference-
>0–5 years	405	0.0 (−9.6, 10.5)	85	−0.6 (−20.9, 24.9)	46	2.7 (−24.4, 39.5)	274	−2.3 (−13.6, 10.5)		449	5.2 (−2.7, 13.7)	54	15.7 (−12.3, 52.6)	53	−1.9 (−22.3, 23.8)	342	4.9 (−4.1, 14.7)
6–10 years	212	3.2 (−8.0, 15.8)	25	−7.2 (−30.6, 24.0)	13	21.0 (−18.3, 79.1)	174	2.7 (−10.3, 17.7)		188	3.2 (−6.5, 13.9)	12	31.3 (−12.2, 96.5)	16	42.0 (1.5, 98.8)	160	−3.7 (−13.8, 7.6)
≥11 years	246	−4.8 (−15.0, 6.5)	15	19.3 (−16.6, 70.6)	24	−0.9 (−30.6, 41.5)	207	−6.9 (−18.3, 6.2)	0.31	141	8.0 (−3.3, 20.7)	5	31.1 (−27.6, 137.4)	14	16.0 (−17.1, 62.4)	122	4.0 (−8.0, 17.6)	0.042
HT usee
Never	-	-	-	-	-	-	-	-		684	-Reference-	78	-Reference-	98	-Reference-	508	-Reference-
Ever	-	-	-	-	-	-	-	-	-	443	−2.3 (−8.9, 4.7)	18	3.7 (−24.1, 41.7)	36	−3.5 (−23.1, 21.1)	389	−1.9 (−9.2, 6.0)	0.60

Abbreviations: NH Non-Hispanic; OC oral contraceptive; HT hormone therapy

^aAll models are adjusted for: age, BMI, alcohol use (0–1 v. >1 drinks per week), smoking (never, past, current), family history of breast cancer (yes, no), history of benign breast disease (ever, never)

^bFurther adjusted for all other reproductive factors, adjustment for HT use only among postmenopausal women

^cFurther adjusted for age at menarche, OC use, and, in postmenopausal women, HT use

^dParous women only: further adjusted for all other reproductive factors, adjustment for HT use only among postmenopausal women

^ePostmenopausal women only: further adjusted for all other reproductive factors

Table 3.

Multivariable-adjusted^a percent difference in dense volume (DV) comparing categories of reproductive factors

	Premenopausal									Postmenopausal
	All premenopausal			Hispanic		NH Black		NH White	p-het	All postmenopausal			Hispanic		NH Black		NH White	p-het
	n	% difference (95% CI)	n	% difference (95% CI)	n	% difference (95% CI)	n	% difference (95% CI)		n	% difference (95% CI)	n	% difference (95% CI)	n	% difference (95% CI)	n	% difference (95% CI)
Age at menarcheb
Age 10 or earlier	46	−12.4 (−27.2, 5.4)	15	−11.5 (−37.4, 25.2)	6	−22.3 (−54.3, 32.2)	25	−5.9 (−27.2, 21.6)		86	−2.9 (−14.4, 10.1)	8	10.4 (−27.7, 68.7)	19	6.6 (−18.1, 38.6)	59	−11.0 (−24.2, 4.3)
Age 11–12	408	-Reference-	64	-Reference-	45	-Reference-	299	-Reference-		439	-Reference-	48	-Reference-	32	-Reference-	359	-Reference-
Age 13 or later	609	3.4 (−3.9, 11.2)	94	2.2 (−14.5, 22.2)	59	22.1 (−1.1, 50.8)	456	2.7 (−5.9, 12.2)	0.89	621	0.7 (−5.4, 7.3)	51	−15.0 (−30.3, 3.7)	86	−1.5 (−18.1, 18.5)	484	3.0 (−4.3, 10.9)	0.36
Parityc
Parous	765	-Reference-	137	-Reference-	77	-Reference-	551	-Reference-		829	-Reference-	78	-Reference-	111	-Reference-	640	-Reference-
Nulliparous	176	0.3 (−8.4, 9.8)	15	15.5 (−15.0, 56.9)	13	38.0 (1.3, 88.0)	148	−3.3 (−12.7, 7.1)	0.041	193	16.4 (7.9, 25.6)	8	−16.2 (−42.2, 21.6)	7	28.5 (−6.9, 77.5)	178	17.1 (7.9, 27.1)	0.20
Number of birthsd
1	165	-Reference-	19	-Reference-	17	-Reference-	129	-Reference-		173	-Reference-	11	-Reference-	22	-Reference-	140	-Reference-
2	332	−12.3 (−20.8, −3.0)	54	−11.7 (−34.9, 19.9)	27	2.0 (−25.4, 39.5)	251	−14.0 (−23.5, −3.3)		330	3.6 (−5.0, 13.1)	20	−2.3 (−34.6, 46.0)	32	−5.2 (−25.5, 20.7)	278	5.7 (−4.3, 16.7)
≥3	268	−14.0 (−23.1, −3.7)	64	−15.2 (−39.2, 18.1)	33	−7.8 (−34.8, 30.4)	171	−13.3 (−24.0, −1.1)	0.73	326	−6.8 (−15.3, 2.6)	47	−17.3 (−43.3, 20.6)	57	−3.1 (−23.9, 23.3)	222	−5.8 (−15.8, 5.5)	0.85
Age at first birthd
<25	147	-Reference-	66	-Reference-	46	-Reference-	35	-Reference-		335	-Reference-	49	-Reference-	83	-Reference-	203	-Reference-
25–29	161	2.5 (−11.7, 18.9)	31	13.6 (−14.0, 50.1)	13	21.7 (−14.7, 73.7)	117	−3.2 (−23.2, 22.1)		233	3.7 (−6.1, 14.5)	17	−1.9 (−29.6, 36.7)	17	11.2 (−15.8, 46.7)	199	3.3 (−8.1, 16.2)
≥30	451	−3.6 (−20.6, 17.0)	36	10.3 (−28.6, 70.4)	18	−25.5 (−56.7, 28.2)	397	−4.4 (−27.2, 25.4)	0.57	256	10.9 (−4.2, 28.4)	12	−8.3 (−45.6, 54.6)	10	13.8 (−26.5, 76.3)	234	14.1 (−3.8, 35.3)	0.12
Time from menarche to first birthd
<15 years	227	-Reference-	84	-Reference-	55	-Reference-	88	-Reference-		455	-Reference-	59	-Reference-	86	-Reference-	310	-Reference-
≥15 years	528	9.5 (−6.9, 28.7)	47	−2.9 (−35.2, 45.7)	21	49.1 (−10.3, 147.9)	460	12.4 (−7.6, 36.8)	0.50	360	−0.8 (−12.0, 11.9)	18	−20.1 (−50.6, 29.2)	21	23.0 (−18.2, 85.1)	321	−2.1 (−14.4, 11.9)	0.047
Breastfeedingd
Parous, did not breastfeed	111	-Reference-	26	-Reference-	14	-Reference-	71	-Reference-		242	-Reference-	15	-Reference-	36	-Reference-	191	-Reference-
>0–6 months	196	2.4 (−9.8, 16.2)	55	−25.0 (−42.9, −1.4)	29	−15.6 (−40.8, 20.4)	112	11.4 (−5.4, 31.3)		210	−0.7 (−9.1, 8.6)	34	19.9 (−14.4, 68.0)	25	−5.7 (−27.6, 22.7)	151	−0.7 (−10.5, 10.3)
>6 months	437	0.6 (−10.5, 13.1)	52	−25.6 (−44.1, −0.9)	32	−7.1 (−36.5, 36.0)	353	9.4 (−5.1, 26.1)	0.13	334	−2.7 (−10.6, 6.0)	22	13.1 (−18.6, 57.1)	36	−5.9 (−26.8, 20.9)	276	−3.2 (−12.3, 6.7)	0.98
OC useb
Never	170	-Reference-	36	-Reference-	23	-Reference-	111	-Reference-		335	-Reference-	31	-Reference-	44	-Reference-	260	-Reference-
>0–5 years	405	3.9 (−6.8, 15.8)	85	1.5 (−18.9, 27.0)	46	−6.9 (−30.0, 23.8)	274	5.4 (−8.4, 21.3)		449	1.5 (−5.8, 9.2)	54	−1.8 (−22.4, 24.1)	53	−4.7 (−20.2, 14.0)	342	4.0 (−4.8, 13.6)
6–10 years	212	−8.1 (−18.8, 4.1)	25	−1.1 (−25.6, 31.5)	13	−25.1 (−48.0, 8.0)	174	−6.5 (−19.9, 9.2)		188	2.9 (−6.3, 13.0)	12	28.3 (−8.8, 80.4)	16	−2.7 (−24.8, 25.9)	160	4.0 (−6.7, 16.0)
≥11 years	246	−10.4 (−20.7, 1.2)	15	11.4 (−21.6, 58.2)	24	−15.5 (−39.3, 17.8)	207	−10.7 (−23.1, 3.8)	0.71	141	2.7 (−7.5, 14.1)	5	−2.1 (−40.8, 61.9)	14	−10.6 (−30.8, 15.7)	122	4.9 (−7.0, 18.4)	0.54
HT usee
Never	-	-	-	-	-	-	-	-		684	-Reference-	78	-Reference-	98	-Reference-	508	-Reference-
Ever	-	-	-	-	-	-	-	-	-	443	2.0 (−4.6, 9.0)	18	6.4 (−18.2, 38.4)	36	17.4 (−2.4, 41.2)	389	−0.2 (−7.5, 7.8)	0.29

Abbreviations: NH Non-Hispanic; OC oral contraceptive; HT hormone therapy

^aAll models are adjusted for: age, BMI, alcohol use (0–1 v. >1 drinks per week), smoking (never, past, current), family history of breast cancer (yes, no), history of benign breast disease (ever, never)

^bFurther adjusted for all other reproductive factors, adjustment for HT use only among postmenopausal women

^cFurther adjusted for age at menarche, OC use, and, in postmenopausal women, HT use

^dParous women only: further adjusted for all other reproductive factors, adjustment for HT use only among postmenopausal women

^ePostmenopausal women only: further adjusted for all other reproductive factors

Reproductive breast cancer risk factors associated with NDV are provided in Table 4. As with PVD and DV, factors most strongly associated with NDV included age at menarche, nulliparity, and number of births among parous women. We observed no evidence of an association between age at menarche and NDV among premenopausal women, but younger age at menarche was positively associated with NDV among postmenopausal women (e.g., 17.6% [3.9% to 33.1%] higher NDV among those with menarche at or before age 10 compared to age 11 or 12). Associations were similar across racial and ethnic subgroups. Nulliparity was positively associated with NDV among premenopausal women (i.e., 11.6% [2.9% to 20.9%] higher NDV among nulliparous women compared to parous women), with the strongest magnitude of association among non-Hispanic Black women (52.8% [2.6% to 127.5%]; p-heterogeneity=0.031). Nulliparity was not associated with NDV among postmenopausal women but, among parous postmenopausal women, those with three or more births had 9.4% (−17.6% to −0.4%) lower NDV compared to those with just one birth.

Table 4.

Multivariable-adjusted^a percent difference in non-dense volume (NDV) comparing categories of reproductive factors

	Premenopausal									Postmenopausal
	All premenopausal			Hispanic		NH Black		NH White	p-het	All postmenopausal			Hispanic		NH Black		NH White	p-het
	n	% difference (95% CI)	n	% difference (95% CI)	n	% difference (95% CI)	n	% difference (95% CI)		n	% difference (95% CI)	n	% difference (95% CI)	n	% difference (95% CI)	n	% difference (95% CI)
Age at menarcheb
Age 10 or earlier	46	−11.7 (−25.1, 4.1)	15	−15.3 (−38.3, 16.2)	6	−33.6 (−64.2, 23.1)	25	3.4 (−17.0, 28.9)		86	17.6 (3.9, 33.1)	8	25.4 (−21.3, 99.8)	19	23.9 (−8.3, 67.5)	59	12.2 (−3.4, 30.3)
Age 11–12	408	-Reference-	64	-Reference-	45	-Reference-	299	-Reference-		439	-Reference-	48	-Reference-	32	-Reference-	359	-Reference-
Age 13 or later	609	−4.9 (−10.9, 1.5)	94	0.2 (−14.8, 18.0)	59	2.5 (−19.8, 31.0)	456	−6.2 (−13.0, 1.2)	0.35	621	0.1 (−5.9, 6.5)	51	−14.7 (−31.4, 6.1)	86	0.3 (−18.8, 23.9)	484	2.3 (−4.5, 9.5)	0.51
Parityc
Parous	765	-Reference-	137	-Reference-	77	-Reference-	551	-Reference-		829	-Reference-	78	-Reference-	111	-Reference-	640	-Reference-
Nulliparous	176	11.6 (2.9, 20.9)	15	10.3 (−15.2, 43.4)	13	52.8 (2.6, 127.5)	148	6.9 (−2.0, 16.7)	0.031	193	2.3 (−4.7, 10.0)	8	−14.3 (−43.3, 29.4)	7	15.7 (−21.9, 71.3)	178	1.5 (−5.7, 9.2)	0.46
Number of birthsd
1	165	-Reference-	19	-Reference-	17	-Reference-	129	-Reference-		173	-Reference-	11	-Reference-	22	-Reference-	140	-Reference-
2	332	−5.3 (−13.3, 3.4)	54	−24.8 (−42.5, −1.5)	27	6.8 (−27.9, 58.2)	251	−4.9 (−13.8, 5.0)		330	−2.2 (−10.3, 6.6)	20	−32.3 (−56.2, 4.6)	32	−10.1 (−32.7, 20.0)	278	1.0 (−8.0, 10.9)
≥3	268	−7.0 (−15.7, 2.6)	64	−25.9 (−44.7, −0.8)	33	−0.5 (−35.6, 53.7)	171	−4.4 (−14.4, 6.8)	0.69	326	−9.4 (−17.6, −0.4)	47	−36.3 (−57.7, −4.0)	57	−0.8 (−25.6, 32.3)	222	−9.7 (−18.7, 0.4)	0.46
Age at first birthd
<25	147	-Reference-	66	-Reference-	46	-Reference-	35	-Reference-		335	-Reference-	49	-Reference-	83	-Reference-	203	-Reference-
25–29	161	8.3 (−4.9, 23.4)	31	−7.2 (−27.5, 18.7)	13	31.1 (−16.1, 104.9)	117	11.7 (−8.3, 35.9)		233	−4.9 (−13.7, 4.9)	17	−8.6 (−36.2, 30.9)	17	5.0 (−24.6, 46.3)	199	−4.9 (−14.8, 6.2)
≥30	451	6.9 (−9.7, 26.6)	36	−6.9 (−36.6, 36.8)	18	3.3 (−47.7, 104.2)	397	13.6 (−9.8, 43.0)	0.24	256	−4.0 (−16.9, 10.9)	12	−41.6 (−66.9, 2.9)	10	−17.1 (−50.9, 39.7)	234	1.2 (−13.8, 18.8)	0.008
Time from menarche to first birthd
<15 years	227	-Reference-	84	-Reference-	55	-Reference-	88	-Reference-		455	-Reference-	59	-Reference-	86	-Reference-	310	-Reference-
≥15 years	528	−5.1 (−17.7, 9.4)	47	−6.1 (−34.4, 34.4)	21	−13.5 (−55.0, 66.4)	460	−4.7 (−19.3, 12.6)	0.20	360	1.1 (−10.1, 13.8)	18	−15.4 (−49.5, 41.7)	21	31.2 (−19.6, 114.2)	321	−1.1 (−12.7, 12.2)	0.010
Breastfeedingd
Parous, did not breastfeed	111	-Reference-	26	-Reference-	14	-Reference-	71	-Reference-		242	-Reference-	15	-Reference-	36	-Reference-	191	-Reference-
>0–6 months	196	−4.9 (−14.8, 6.1)	55	4.5 (−17.8, 33.0)	29	−27.2 (−54.0, 15.0)	112	−2.8 (−15.4, 11.6)		210	0.1 (−8.2, 9.2)	34	17.2 (−22.4, 77.2)	25	4.3 (−23.7, 42.4)	151	−3.7 (−12.6, 6.1)
>6 months	437	−7.7 (−16.7, 2.1)	52	1.2 (−21.4, 30.3)	32	−22.9 (−52.8, 26.1)	353	−6.4 (−17.0, 5.5)	0.51	334	−6.4 (−13.9, 1.8)	22	8.2 (−27.7, 61.9)	36	−6.0 (−30.1, 26.4)	276	−8.5 (−16.4, 0.3)	0.28
OC useb	.
Never	170	-Reference-	36	-Reference-	23	-Reference-	111	-Reference-		335	-Reference-	31	-Reference-	44	-Reference-	260	-Reference-
>0–5 years	405	4.4 (−5.1, 15.0)	85	2.7 (−15.8, 25.3)	46	−9.7 (−35.2, 25.7)	274	8.9 (−3.4, 22.7)		449	−3.8 (−10.6, 3.6)	54	−16.1 (−35.7, 9.5)	53	−2.3 (−21.2, 21.0)	342	−0.8 (−8.7, 7.7)
6–10 years	212	−11.1 (−20.4, −0.8)	25	6.5 (−17.2, 37.1)	13	−39.3 (−60.3, −7.2)	174	−8.9 (−20.2, 4.0)		188	−0.5 (−9.4, 9.2)	12	−3.6 (−34.6, 42.1)	16	−34.3 (−51.8, −10.5)	160	8.6 (−1.9, 20.3)
≥11 years	246	−5.2 (−14.9, 5.6)	15	−9.6 (−33.8, 23.3)	24	−14.9 (−42.1, 25.1)	207	−2.7 (−14.4, 10.6)	0.15	141	−5.1 (−14.5, 5.3)	5	−25.8 (−58.1, 31.4)	14	−23.1 (−43.5, 4.7)	122	0.8 (−10.0, 12.9)	0.013
HT usee
Never	-	-	-	-	-	-	-	-		684	-Reference-	78	-Reference-	98	-Reference-	508	-Reference-
Ever	-	-	-	-	-	-	-	-	-	443	4.7 (−1.9, 11.8)	18	1.8 (−24.2, 36.7)	36	22.2 (−0.6, 50.3)	389	2.1 (−4.9, 9.7)	0.86

Abbreviations: NH Non-Hispanic; OC oral contraceptive; HT hormone therapy

^aAll models are adjusted for: age, BMI, alcohol use (0–1 v. >1 drinks per week), smoking (never, past, current), family history of breast cancer (yes, no), history of benign breast disease (ever, never)

^bFurther adjusted for all other reproductive factors, adjustment for HT use only among postmenopausal women

^cFurther adjusted for age at menarche, OC use, and, in postmenopausal women, HT use

^dParous women only: further adjusted for all other reproductive factors, adjustment for HT use only among postmenopausal women

^ePostmenopausal women only: further adjusted for all other reproductive factors

OC use was also inversely associated with NDV (Table 4), particularly among premenopausal women (e.g., 11.1% [−20.4% to −0.8%] lower NDV among women taking OCs for 6–10 years compared with never users of OCs), though increasing OC use was not consistently associated with lower NDV. For example, despite the positive association with NDV when comparing women taking OCs for 6–10 years compared to never users of OCs, premenopausal women taking OCs for ≥11 years had a non-statistically significant 5.2% [−14.9% to 5.6%] lower NDV compared with never users. Heterogeneity in the association between OC use and NDV was not evident among premenopausal women, but was possible among postmenopausal women (p-heterogeneity=0.013). This heterogeneity reflects the inverse association between greater OC use and NDV among Hispanic and non-Hispanic black women, but not among non-Hispanic White women.

None of the reproductive factors were consistently associated with V (Table 5; Supplementary Figure 2). However, we observed racial and ethnic heterogeneity for some associations between reproductive factors and V among postmenopausal women. For example, we observed variation in the strength and direction of associations between the number of births and V (p=0.014), with positive associations among non-Hispanic Black postmenopausal women (e.g., β=0.3 [−0.1 to 0.7] comparing 3 births to 1 birth), inverse associations among Hispanic postmenopausal women (e.g., β=−0.4 [−1.1 to 0.3] comparing 3 births to 1 birth), and no evidence of association among non-Hispanic White postmenopausal women (e.g., β=0.0, [−0.1 to 0.2] comparing 3 births to 1 birth). Associations between time from menarche to first birth and V also varied among postmenopausal women (p=0.004), though confidence intervals were wide for the Hispanic and non-Hispanic Black groups (Table 5).

Table 5.

Multivariable-adjusted^a difference in V-metric comparing categories of reproductive factors

	Premenopausal									Postmenopausal
	All premenopausal			Hispanic		NH Black		NH White	p-het	All postmenopausal			Hispanic		NH Black		NH White	p-het
	n	β (95% CI)	n	β (95% CI)	n	β (95% CI)	n	β (95% CI)		n	β (95% CI)	n	β (95% CI)	n	β (95% CI)	n	β (95% CI)
Age at menarcheb
Age 10 or earlier	46	0.0 (−0.2, 0.2)	15	0.1 (−0.3, 0.4)	6	0.4 (−0.3, 1.1)	25	−0.1 (−0.4, 0.2)		86	0.1 (−0.1, 0.3)	8	0.3 (−0.5, 1.0)	19	0.1 (−0.3, 0.5)	59	0.1 (−0.1, 0.3)
Age 11–12	408	-Reference-	64	-Reference-	45	-Reference-	299	-Reference-		439	-Reference-	48	-Reference-	32	-Reference-	359	-Reference-
Age 13 or later	609	0.0 (−0.1, 0.1)	94	0.0 (−0.2, 0.2)	59	0.3 (0.0, 0.5)	456	0.0 (−0.1, 0.1)	0.65	621	0.0 (−0.1, 0.1)	51	−0.4 (−0.7, 0.0)	86	−0.1 (−0.4, 0.2)	484	0.0 (0.0, 0.1)	0.22
Parityc
Parous	765	-Reference-	137	-Reference-	77	-Reference-	551	-Reference-		829	-Reference-	78	-Reference-	111	-Reference-	640	-Reference-
Nulliparous	176	0.1 (0.0, 0.2)	15	0.2 (−0.2, 0.5)	13	0.0 (−0.5, 0.5)	148	0.1 (−0.1, 0.2)	0.71	193	0.0 (−0.1, 0.1)	8	−0.1 (−0.7, 0.6)	7	0.1 (−0.5, 0.6)	178	0.1 (−0.0, 0.2)	0.84
Number of birthsd
1	165	-Reference-	19	-Reference-	17	-Reference-	129	-Reference-		173	-Reference-	11	-Reference-	22	-Reference-	140	-Reference-
2	332	−0.1 (−0.2, 0.1)	54	0.1 (−0.2, 0.4)	27	0.1 (−0.3, 0.5)	251	−0.1 (−0.2, 0.0)		330	0.1 (0.0, 0.2)	20	−0.1 (−0.9, 0.6)	32	0.7 (0.3, 1.1)	278	0.1 (−0.1, 0.2)
≥3	268	0.0 (−0.2, 0.1)	64	0.4 (0.0, 0.7)	33	0.3 (−0.2, 0.7)	171	−0.1 (−0.3, 0.0)	0.16	326	0.0 (−0.1, 0.2)	47	−0.4 (−1.1, 0.3)	57	0.3 (−0.1, 0.7)	222	0.0 (−0.1, 0.2)	0.014
Age at first birthd
<25	147	-Reference-	66	-Reference-	46	-Reference-	35	-Reference-		335	-Reference-	49	-Reference-	83	-Reference-	203	-Reference-
25–29	161	0.0 (−0.2, 0.1)	31	0.2 (−0.1, 0.5)	13	−0.1 (−0.5, 0.4)	117	−0.1 (−0.4, 0.2)		233	0.1 (0.0, 0.3)	17	−0.5 (−1.1, 0.1)	17	0.5 (0.1, 1.0)	199	0.1 (−0.1, 0.2)
≥30	451	−0.1 (−0.4, 0.1)	36	0.2 (−0.3, 0.7)	18	−0.2 (−0.9, 0.6)	397	−0.2 (−0.5, 0.2)	0.95	256	0.1 (−0.1, 0.4)	12	0.0 (−0.9, 1.0)	10	0.6 (−0.1, 1.4)	234	0.1 (−0.1, 0.3)	0.22
Time from menarche to first birthd
<15 years	227	-Reference-	84	-Reference-	55	-Reference-	88	-Reference-		455	-Reference-	59	-Reference-	86	-Reference-	310	-Reference-
≥15 years	528	0.1 (−0.1, 0.3)	47	0.4 (−0.1, 0.8)	21	0.3 (−0.4, 1.0)	460	0.1 (−0.1, 0.3)	0.84	360	0.0 (−0.2, 0.2)	18	0.0 (−0.9, 0.9)	21	0.2 (−0.5, 0.9)	321	−0.1 (−0.2, 0.1)	0.004
Breastfeedingd
Parous, did not breastfeed	111	-Reference-	26	-Reference-	14	-Reference-	71	-Reference-		242	-Reference-	15	-Reference-	36	-Reference-	191	-Reference-
>0–6 months	196	0.1 (0.0, 0.2)	55	−0.1 (−0.3, 0.2)	29	0.1 (−0.4, 0.5)	112	0.1 (−0.1, 0.3)		210	0.0 (−0.1, 0.1)	34	0.4 (−0.3, 1.0)	25	0.0 (−0.4, 0.5)	151	0.0 (−0.1, 0.1)
>6 months	437	0.0 (−0.1, 0.2)	52	−0.3 (−0.6, 0.0)	32	−0.2 (−0.7, 0.3)	353	0.1 (−0.1, 0.3)	0.11	334	−0.1 (−0.3, 0.0)	22	−0.1 (−0.7, 0.6)	36	−0.4 (−0.8, 0.0)	276	−0.1 (−0.2, 0.0)	0.81
OC useb
Never	170	-Reference-	36	-Reference-	23	-Reference-	111	-Reference-		335	-Reference-	31	-Reference-	44	-Reference-	260	-Reference-
>0–5 years	405	−0.1 (−0.2, 0.1)	85	−0.1 (−0.4, 0.1)	46	0.1 (−0.3, 0.5)	274	−0.1 (−0.2, 0.1)		449	−0.1 (−0.2, 0.0)	54	−0.2 (−0.7, 0.2)	53	−0.2 (−0.5, 0.1)	342	−0.1 (−0.2, 0.0)
6–10 years	212	−0.1 (−0.3, 0.0)	25	−0.1 (−0.5, 0.2)	13	−0.2 (−0.7, 0.3)	174	−0.1 (−0.3, 0.1)		188	0.0 (−0.1, 0.1)	12	−0.2 (−0.8, 0.5)	16	0.0 (−0.4, 0.5)	160	0.0 (−0.2, 0.1)
≥11 years	246	−0.1 (−0.3, 0.0)	15	−0.2 (−0.6, 0.2)	24	−0.3 (−0.8, 0.1)	207	−0.1 (−0.3, 0.1)	0.45	141	0.1 (0.0, 0.3)	5	−0.5 (−1.4, 0.5)	14	0.3 (−0.2, 0.7)	122	0.1 (−0.1, 0.2)	0.42
HT usee
Never	-	-	-	-	-	-	-	-		684	-Reference-	78	-Reference-	98	-Reference-	508	-Reference-
Ever	-	-	-	-	-	-	-	-	-	443	0.0 (−0.1, 0.1)	18	−0.3 (−0.8, 0.2)	36	0.0 (−0.3, 0.2)	389	0.0 (−0.1, 0.1)	0.82

^aAll models are adjusted for: age, BMI, alcohol use (0–1 v. >1 drinks per week), smoking (never, past, current), family history of breast cancer (yes, no), history of benign breast disease (ever, never), percent volumetric density

^bFurther adjusted for all other reproductive factors, adjustment for HT use only among postmenopausal women

^cFurther adjusted for age at menarche, OC use, and, in postmenopausal women, HT use

^dParous women only: further adjusted for all other reproductive factors, adjustment for HT use only among postmenopausal women

^ePostmenopausal women only: further adjusted for all other reproductive factors

To ensure comparability with studies that measured breast density using area-based density metrics, we re-ran analyses using LIBRA area-based metrics (Supplemental Tables 1–4). While magnitudes of association varied, patterns in the direction and statistical significance of associations remained similar.

DISCUSSION

In this multiracial and multiethnic clinical cohort, we observed variation in the distribution of BD measures and reproductive breast cancer risk factors by race and ethnicity. We also reported evidence of racial and ethnic heterogeneity in the associations between reproductive factors and these measures. Our finding of racial and ethnic variation in the prevalence of BD features and reproductive factors adds to evidence from prior studies that report racial and ethnic variation in the distribution of breast cancer risk factors [3–9]. For example, one study of BD reported racial and ethnic variation in BMI, parity, age at first birth, and use of postmenopausal HT [9], and a number of breast cancer studies reported racial and ethnic variation in the distributions of BD [8], BMI [7, 8], parity [3, 5, 7], age at first full-term pregnancy [3, 7], breastfeeding duration [5, 7], and use of postmenopausal HT [7]. In aggregate, these studies suggest that interventions focusing on factors that can be modified on an individual or population level could help to reduce breast cancer disparities.

The extent to which associations between reproductive factors and BD vary across populations is not fully understood. In the present study, we observed the greatest racial and ethnic heterogeneity in associations between parity and both DV and NDV among premenopausal women, and between duration of OC use and both PVD and NDV in postmenopausal women. Heterogeneity was also evident in the associations between number of births and V, and breastfeeding duration and V.

To our knowledge, this is the first study to directly compare the associations between parity and PVD, DV and NDV by race and ethnicity among premenopausal women. We reported the strongest positive association between nulliparity and DV among non-Hispanic Black women, a weaker positive association for Hispanic women, and no association for non-Hispanic White women. The positive association between nulliparity and NDV was also strongest in non-Hispanic Black women and weaker among Hispanic and non-Hispanic White women. Prior studies among premenopausal Hispanic and non-Hispanic White women produced findings similar to ours with no statistically significant associations between nulliparity and PD or PVD, a mix of positive and inverse associations between nulliparity and DA or DV, and a mix of inverse and null associations between nulliparity and NDA or NDV [26, 28].

To our knowledge, this is also the first study to directly compare the associations between OC use and PVD, DV and NDV by race and ethnicity among postmenopausal women. We observed the strongest positive association with PVD among Hispanic women and the weakest association among non-Hispanic White women. This appeared to be driven by a strong inverse association between OC use and NDV among Hispanic women and, to a lesser extent, non-Hispanic Black women, but not non-Hispanic White women. Interestingly, despite the strongest associations in our study occurring among Hispanic women, a prior study of OC use and BD in the Mexican Teachers’ Cohort (MTC) reported no statistically significant association between OC use and PD, DA, or NDA [26]. This difference across studies could be attributable to differences in the primary outcome metric, categorization of the exposure, nativity, or ethnic origins of populations, or chance.

V is a novel BD feature that is associated with breast cancer risk independent of both DV and PD [42, 43]. Prior studies have reported associations between anthropometric factors and V [41], but the associations between reproductive factors and V have not been studied. Findings from this study that would especially benefit from replication and etiologic evaluation include (1) the relatively consistent inverse associations between longer duration of breastfeeding and lower V, and (2) the racial and ethnic heterogeneity in the associations between number of births and V, and breastfeeding duration and V.

The primary strength of our approach to characterizing BD and evaluating how reproductive factors are associated with BD was the use of mammography unit-agnostic, automated approaches to measure BD. The consideration of both automated volumetric measures (PVD, DV, and NDV) and V allowed us to detect differences that may not have been captured in an ordinal measurement such as the Breast Imaging Reporting and Database System (BI-RADS) breast composition categories [47]. In this study, we collected self-reported race and ethnicity and menopausal status at the time of mammography, which was also a design strength, allowing for stratified analyses and assessments of heterogeneity. While our study was multiracial and multiethnic, a limitation was the small number of non-White women. We know from previous research that the distribution of BD features differs across racial and ethnic groups, yet we only were able to include Hispanic, non-Hispanic Black, and non-Hispanic White populations, and confidence intervals were wide for some comparisons [8, 9, 29, 48–50]. Future studies in larger, more diverse populations are warranted.

Taken together, the findings from this study show variation in the distributions of reproductive risk factors for breast cancer and BD by race and ethnicity, but limited racial and ethnic heterogeneity in how these same reproductive factors are associated with BD measures. If future research is able to describe how reproductive factors influence the biology of breast tissue, especially features easily extracted from a mammogram, it may be possible to refine risk prediction and screening approaches for women with specific reproductive histories—making them more effective and more equitable.

Data Availability

Study data may be shared upon reasonable request and with approval of the appropriate Institutional Review Boards.