Research on Health Disparities: Strategies and Findings From the Black Women’s Health Study

Abstract

The American Journal of Epidemiology has been a platform for findings from the Black Women’s Health Study (BWHS) that are relevant to health disparities. Topics addressed have included methods of follow-up of a large cohort of Black women, disparities in health-care delivery, modifiable risk factors for health conditions that disproportionately affect Black women, associations with exposures that are highly prevalent in Black women, and methods for genetic research. BWHS papers have also highlighted the importance of considering social context, including perceived experiences of racism, in understanding health disparities. In the future, BWHS investigators will contribute to documentation of the role that structural racism plays in health disparities.

Abbreviations

AJE

American Journal of Epidemiology

BMI

body mass index

BWHS

Black Women’s Health Study

ER

estrogen receptor

SES

socioeconomic status

SNP

single nucleotide polymorphism

Editor’s note: The opinions expressed in this article are those of the authors and do not necessarily reflect the views of the American Journal of Epidemiology.

The Black Women’s Health Study (BWHS), a follow-up study of US Black women, was begun in 1995, with funding from the National Cancer Institute (1). The primary impetus for the study was the 40% higher mortality from breast cancer experienced by US Black women relative to US White women (2), as well as a long list of other health conditions, such as stroke, heart disease, type 2 diabetes, and uterine leiomyoma, that disproportionately affect Black women (3, 4). We envisioned a large prospective study that would permit investigation of multiple diseases and exposures. The study design and ongoing decisions about study methods have been informed by our understanding that race is a socially constructed variable which captures neither the complexity of individual genetic risk nor the myriad social determinants that influence health (5, 6). In the United States, both interpersonal and structural racism, rather than genetic differences, are fundamental causes of health disparities (6–8).

To enroll study participants, we mailed invitations to hundreds of thousands of subscribers to Essence magazine, a popular magazine targeted to Black women, inviting participation in a long-term study of health and completion of a questionnaire on demographic characteristics, reproductive history, medical history, and diet. The 59,000 Black women aged 21–69 years who successfully completed and returned the questionnaire comprise the largest cohort of Black women yet studied. They have now been followed for over 25 years with biennial questionnaires (1, 9, 10).

A major challenge for most prospective cohort studies is to successfully follow participants over time with a high rate of response to each questionnaire cycle. In 2001, the American Journal of Epidemiology (AJE) published a report (9) that documented the high proportion of address changes among BWHS participants, especially among younger women, and described measures we took to learn of the changes in a timely fashion. The primary methods were 1) mailing BWHS newsletters every 6 months and thereby obtaining address changes from the US Postal Service for undeliverable newsletters and 2) searching for address changes through commercial databases. The newsletters also had the important role of “giving back” to study participants by providing useful information from both study publications and other sources. An AJE paper published in 2010 (10) described the provision of an online option for completing the biennial health questionnaire. These were early days for use of the Internet in research, and the paper documented the feasibility and acceptability of the approach. The online option, in addition to providing ease of completion for women who preferred it, also provided cost savings, with reductions in expenses for printing and mailing questionnaires, as well as for return postage, for completed questionnaires. In the first questionnaire cycle that offered an online option, 10% of respondents used that mode of response. The proportion opting for online completion has increased in each cycle since then, and in the most recent completed questionnaire cycle, 2019–2020, half of all respondents filled out their questionnaire online.

The initial scientific report from the BWHS addressing racial disparity in health-care delivery was published in the AJE in 1999 (11): BWHS participants were found to have a high prevalence of hysterectomy, consistent with US data indicating that the prevalence of hysterectomy was about 30% higher in Black women than in non-Hispanic White women. Importantly, our data showed that the prevalence of hysterectomy in the BWHS varied by geographic region, with 2.6 times’ higher odds of hysterectomy among women residing in the South compared with women residing in the Northeast. The prevalence was also significantly higher among women whose highest level of educational attainment was high school graduation than among college graduates. These findings suggested that modifiable factors, such as clinician practices and attitudes, were contributors to the high prevalence of hysterectomy in Black women.

BWHS papers published in the AJE have addressed health conditions that disproportionately affect Black women, with an emphasis on modifiable factors. With regard to type 2 diabetes, we reported that women who engaged in 5 or more hours of vigorous physical activity each week had half the risk of type 2 diabetes as women who did not exercise, and the association was consistent across levels of body mass index (BMI; weight (kg)/height (m)²) (12). In addition, the risk of diabetes increased as hours of television-watching increased, even among women who exercised regularly. Because the incidence of uterine leiomyoma is 2–3 times higher in Black women than in White women, we have conducted numerous analyses of potential risk factors for this condition, including AJE papers on oral contraceptive use (13), dietary factors (14, 15), and use of hair relaxers (16); none of these analyses identified factors that could explain the large racial difference in incidence. Regarding breast cancer risk (17), we found that women who ate 2 or more servings of vegetables per day, relative to those who ate 4 or fewer servings per week, had approximately half the risk of developing estrogen-receptor (ER)-negative breast cancer, a poor-prognosis breast cancer subtype that occurs twice as often in Black women as in all other US women; vegetable intake was not associated with risk of ER-positive breast cancer. These results were later confirmed in a large pooled analysis of data from 20 studies (18).

Another approach to understanding racial disparities in disease occurrence is to focus on exposures that differ in prevalence between groups. Among these factors are obesity and breastfeeding. In a 2016 AJE paper (19), we reported on the association of obesity with endometrial cancer risk. Because of the high prevalence of obesity among US Black women, we were able to obtain stable estimates of relative risk for women in the highest levels of BMI relative to women with BMI less than 25, finding an incremental increase with increasing BMI and an estimated rate ratio of 3.60 for women with BMI greater than or equal to 40. Until recently, US incidence rates for endometrial cancer were lower for Black women relative to other groups, a pattern that was probably explained by the higher prevalence of premenopausal hysterectomy in Black women (11). A reduction of unnecessary hysterectomies in the Black population could result in a Black/White reversal in relative incidence rates of endometrial cancer, due at least in part to the higher prevalence of excess weight and adverse metabolic indices in Black women (20). With regard to breastfeeding, despite efforts to promote breastfeeding among all new mothers in the United States, the prevalence among parous Black women remains substantially lower than in other groups (approximately 50% in Black women vs. approximately 80% in others) (21). A complex set of factors probably accounts for this difference, including the legacy of slavery, in which many Black women were forced to serve as “wet nurses” for other women’s babies, economic conditions that prohibit time off from work in the early months after childbirth, lack of facilities for pumping at many workplaces, cultural norms, and experiences of racism (22–25). In 2 AJE papers from the BWHS on breastfeeding, we showed that breastfeeding has health benefits for the mother herself: Women who breastfed for at least 6 months had lower long-term weight gain (26) and were less likely to develop hypertension (27).

Race and socioeconomic status (SES) are often interlinked, and it has been difficult to know whether differences in disease incidence by race are primarily due to SES differences. The participants in the BWHS live in a wide range of neighborhoods, and many women of high personal SES live in poorer neighborhoods, a reflection of the legacy of redlining and similar practices limiting where Black Americans can live (7, 28). We have evaluated disease incidence by personal and area-level SES and examined associations of other factors within strata of SES. In our 2010 AJE paper on the relationship of neighborhood SES to risk of type 2 diabetes (29), we reported that women who lived in the US Census block groups that ranked in the lowest quintile of a neighborhood SES score based on 6 census variables had a 65% increased risk of type 2 diabetes compared with women who lived in neighborhoods that fell within the highest quintile. Importantly, this association was present even among women with the highest levels of education and income. Thus, where one lives can have an impact over and above an individual’s personal SES. We applied the same methods to research on breast cancer etiology. As reported in a 2016 AJE paper (30), women who lived in the highest quintile of neighborhood SES were estimated to have a reduced risk of ER-negative breast cancer, although the trend across neighborhood SES was not statistically significant; neighborhood SES was not associated with risk of ER-positive breast cancer. These results are in line with data from US studies indicating higher incidence of ER-negative breast cancer in areas characterized by low SES (31–33).

At the other end of the spectrum from research on area-level factors is research into germline genetic susceptibility. With the completion of the Human Genome Project and other advances that revolutionized research on genetic susceptibility, we collected saliva samples from BWHS participants to be used as a source of germline DNA. During 2003–2007, approximately 50% of study participants provided a sample (34). BWHS papers on associations of genetic variants with specific health conditions have generally been published in disease-specific or genetic journals (35–41). These contributions are important, not because genetic differences across ancestral populations explain observed disparities in health, but because the benefits derived from genetic research will not reach underrepresented populations if most genetic research continues to be carried out in populations of European ancestry. Due to differences in linkage disequilibrium and the substantially greater genetic diversity in populations of African ancestry, variants associated with disease and disease processes in other populations often do not transfer well to African-ancestry populations (42, 43). Thus, for example, polygenic risk scores, which have value in distinguishing individuals at low versus high risk of a given disease, will not perform satisfactorily in an African-ancestry population until variants derived from African-ancestry or multiethnic genome-wide association studies are included (44, 45). As of 2018, 79% of participants with genome-wide association study data were of European ancestry, with African-ancestry individuals representing only a tiny proportion, and the disparity in proportions was increasing over time rather than decreasing (44). Samples from large studies of African or African-American participants, such as the BWHS, are critical for remedying this situation.

A few BWHS genetic papers of wider interest to the epidemiology community were published in AJE. One of the earliest involved ancestry-informative markers, genetic variants that show large allele-frequency differences among the ancestral populations from which admixed populations have originated. In our 2011 AJE publication, we reported that a small set of ancestry-informative markers could be used for accurate estimation of global percentage of African versus European ancestry (46), thus obviating the need for including thousands of single nucleotide polymorphisms (SNPs) on genotyping platforms for this purpose. At the time of publication, genome-wide SNP arrays were expensive, and most genotyping for epidemiologic research was being conducted on smaller arrays for which cost was calculated on the basis of number of SNPs. Based on the small set of ancestry-informative markers, BWHS participants with a higher proportion of African ancestry were estimated to have a higher risk of uterine leiomyoma than participants with lower proportions (47). We used this information to follow up on an earlier AJE publication (14) in which we reported an inverse association between dairy food intake and risk of uterine leiomyoma. Given that dairy food intake is lower in individuals who have lactose intolerance, that lactose intolerance is more common in US Black women than in US White women, and that higher levels of African admixture were associated with increased risk of uterine leiomyoma, we wondered whether the dairy food–leiomyomata association was confounded by ancestry. We repeated the original analysis with additional adjustment for percentage of African (versus European) ancestry and found the associations to be essentially unchanged (48). That paper was named an “Article of the Year for 2013” by the AJE and the Society for Epidemiologic Research (49). In another AJE paper (50), we reported that higher proportions of African ancestry (relative to European ancestry) were associated with lower serum levels of vitamin D in the BWHS, suggesting that factors related to genetic ancestry, including genes related to skin pigmentation and metabolism, are likely to contribute to interpersonal variability in vitamin D levels.

Despite the contributions of genetic analyses to understanding disease etiology, genetics and/or biology are not dominant influences on health disparities. More important are social determinants of health, including experiences of racism (7, 51, 52). Experiences of interpersonal racism have been shown to be associated with biological stress and disruption of the body’s immune, neuroendocrine, and autonomic systems (53). In a BWHS study of racism and obesity, published in the AJE (54), higher levels of perceived experiences of racism relative to low levels were associated with greater risk of obesity. Experiences of racism were also associated in the BWHS with increased risk of a wide variety of other adverse outcomes, including preterm birth (55), uterine leiomyomata (56), weight gain (57), adult-onset asthma (58), diabetes (59), insomnia (60), and reduced cognitive function (61).

Beyond perceived experiences of racism, “structural” racism may be an important driver of racial health disparities (8). Structural racism includes the many different ways in which societies foster racial discrimination through policies on housing, education, employment, incarceration, and health care and other means (52). Among the earliest studies of structural racism were those showing that “redlining”—a practice preventing Black Americans from living in neighborhoods of higher SES, thus denying them access to healthy food, green space, and other contributors to good health—resulted in poorer health among generations affected by it (62). Analyses of the effects of structural racism on the health of BWHS participants are in progress.

In summary, the AJE has been a key platform for BWHS publications that may interest a wide readership of the epidemiologic community. The publications have described design strategies, including a focus on within-group comparisons rather than comparisons across populations, disparities in health-care delivery, findings on modifiable factors that impact health conditions that disproportionately affect Black women, associations with exposures that are highly prevalent in Black women, and the importance of considering social context in addition to individual-level factors. The study has demonstrated adverse influences of perceived experiences of racism on a range of health outcomes. In the past few years, we have seen increased appreciation by the scientific community of the importance of research on health disparities, especially research that moves beyond describing disparities to research that identifies factors that can be modified to reduce disparities. It is our hope that the experience of the first 25 years of the BWHS has illustrated informative approaches to racial health disparities research and will aid the next generation of disparities researchers to improve on what has gone before. We also expect that both the AJE and the BWHS will be in the forefront of understanding how structural racism contributes to racial health disparities. It is to be hoped that documentation of the contribution of structural racism to health disparities will increase support for dismantling these systems.