|Home | About | Journals | Submit | Contact Us | Français|
Current evidence on breast cancer among US Hispanic women indicates a significant public health threat, although few studies assess the heterogeneity in breast cancer risk among Hispanics of different origin.
The 2000 and 2005 National Health Interview Survey (NHIS) Cancer Control Modules were used to examine the Breast Cancer Risk Assessment Tool (BCRAT) 5-year and lifetime risk of invasive breast cancer among Mexican/Mexican-American, Puerto Rican, Cuban/Cuban-American, Dominican (Republic), Central/South American, Other Hispanic and non-Hispanic white (NHW) women aged 35-84 years. Multiple linear regression models were used to compare the BCRAT 5-year and lifetime breast cancer risk between: i) Hispanics and NHWs and ii) Hispanic subgroups.
Hispanics had significantly lower mean BCRAT 5-year and lifetime breast cancer risk compared to NHWs (p<0.001). Among Hispanic subgroups, Cuban/Cuban-Americans had a higher BCRAT 5-year risk (p<0.05), while Dominicans had a higher lifetime risk (p<0.001), compared to Mexican/Mexican-American women. Approximately, 2.6% of Hispanic women were at high-risk for breast cancer (BCRAT 5-year risk ≥1.67%), ranging from 1.0% of Central/South Americans to 3.7% of Puerto Ricans; few Hispanics (0.2%) had a lifetime risk ≥20.0%.
Our findings indicate that Hispanics had significantly lower risk of breast cancer, compared to NHWs, though BCRAT risk significantly differed between specific Hispanic subgroups. We provide estimates of the number of US Hispanic women, from six subgroups, who would be eligible for prophylactic breast cancer chemoprevention. Future studies should further investigate the heterogeneity in breast cancer risk and risk factors between Hispanic women of different origins.
Breast cancer is a significant public health threat to Hispanic women in the United States (US), as it represents the most frequently diagnosed cancer and leading cause of cancer-related death in this population1. Although recent trends show declining incidence of breast cancer among US women, breast cancer incidence has declined at a slower rate for US Hispanic women2; furthermore, these findings are consistent across stages of diagnosis, as the rate at which large breast cancer tumors are diagnosed among Hispanic women has not significantly declined3. Despite having a low incidence rate of breast cancer, Hispanic women are more likely to present with advanced breast cancer at diagnosis4, 5 and more likely to die from breast cancer compared non-Hispanic white women6.
Although these data reflect the burden of breast cancer among Hispanic women as a whole, it is also important to understand the diversity of this population and the extent to which the risk of breast cancer varies between Hispanic women of different national origins and backgrounds. US Hispanics women are characterized by differences in genetic ancestry, behavioral and lifestyle practices, which may lead to differences in the risk of developing breast cancer2. However, the heterogeneity in breast cancer risk among women in different Hispanic subgroups has not been well-defined.
To address this gap, our objective was to assess the distribution of breast cancer risk among Hispanic women from six subgroups: Mexican/Mexican-American, Puerto Rican, Cuban/Cuban-American, Dominican (Republic), Central/South American, and Other Hispanic. Using data from the 2000 and 2005 National Health Interview Survey (NHIS) Cancer Control Modules, we assessed the 5-year and lifetime risk of developing invasive breast cancer among US Hispanic (n=3,386) and non-Hispanic white (n=16,131) women, based on the National Cancer Institute (NCI) Breast Cancer Risk Assessment Tool (BCRAT). Evaluating breast cancer risk among different Hispanic subgroups has the potential to improve our understanding of the impact of breast cancer in this diverse population, and to identify those women from specific Hispanic subgroup who could benefit from breast cancer risk reduction strategies.
Information about the study design, data source, and study population has been previously described in detail7. In brief, the NHIS is an annual, cross-sectional household survey that obtains information on the health of the civilian, non-institutionalized population residing in the United States8. The NHIS is a multistage cluster probability sample survey design, and oversamples both Hispanic and Black persons. Further, the NHIS core questionnaire is comprised of four components: household, family, sample adult, and sample child, with additional supplement questionnaires on specific topics, including: cancer control. The Cancer Control Module (CCM), designed and funded by the National Cancer Institute (NCI), was administered in 2000 and 2005 and it collected information on diet and nutrition, physical activity, tobacco usage, cancer screening, genetic testing, family history and other risk factors related to cancer. The 2000 and 2005 NHIS CCM collected data from one randomly sampled adult 18 years and older, from each sampled family and household, resulting in 32,374 and 31,321 sampled adults, respectively. For this study, we analyzed data on Hispanic (n=3,386) and non-Hispanic white females (n=16,131) aged 35-84 years, with no history of breast cancer or mastectomy (unilateral or bilateral), who completed the CCM module.
Absolute risk is defined as the probability that a person with a given set of risk factors and free of the disease of interest at age x (e.g. age 40 years old), will develop disease before a subsequent age x+y (e.g. age 45 years old), where y is the time interval over which risk is projected9 (e.g. 5 years). The NCI Breast Cancer Risk Assessment Tool10,11 (http://www.cancer.gov/bcrisktool/) estimates a woman’s absolute risk of developing invasive breast cancer over a specific time period based on age, age at first live birth, age at menarche, number of first-degree relatives with breast cancer, number of breast biopsies, and presence of atypical hyperplasia. Information on atypical hyperplasia was unavailable in the NHIS. When information on a particular risk factor is missing, BCRAT imputes the lowest category of risk. We used the BCRAT to estimate participants’ absolute risk of invasive breast cancer over two time periods: i) 5-year risk: from age at interview to the hypothetical age that a woman would attain if she survived 5 years from the date of the interview; and ii) lifetime risk: from age at interview to the hypothetical age that a woman would attain if she survived to age 90 years.
The 2000 and 2005 NHIS collected self-reported data on race and Hispanic origin (based on country of origin and ancestry), which we used to categorize women as non-Hispanic white or Hispanic. Hispanic women were further divided into six distinct subgroups: Mexican/Mexican-American, Cuban/Cuban-American, Puerto Rican, Dominican (Republic), Central or South American, Other Latin American/Other Spanish/Multiple Hispanic (referred to as “Other Hispanic”).
We included several variables in the analyses that previous research indicates may be associated with the risk of developing breast cancer12-20. Sociodemographic variables included marital status (married/living with intimate partner, other), education (less than high school graduate, high school graduate/GED, more than high school), and federal poverty level (<100%, 100-200%, ≥200%). Two variables were included as proxies for access to health care: usual source of care (yes, no) and insurance status (private, public, uninsured). We also included information on county of birth (US born, foreign born), years in the US (<5 years, 5-9 years, ≥10 years) and language most often spoken (mostly/only Spanish, Spanish/English about the same, mostly/only English), and body mass index (BMI<25 “normal”, BMI 25-30 “overweight”, BMI>30 “obese”).
Data from the 2000 and 2005 NHIS were pooled for this analysis. To obtain U.S. population estimates, the observations were weighted by the sample weights for each year, which were summed and divided by two21. The stratified, multi-stage cluster complex sample design of the NHIS was accounted for when calculating standard error (SE) and 95% confidence intervals (CIs)21.
Descriptive statistics were used to assess participants’ baseline sociodemographic, access, acculturation and BMI characteristics. We estimated participants’ BCRAT five-year and lifetime risk estimates, and corresponding 95% CIs as previously described10, 11. Statistical methods appropriate for complex samples were used to compare the distribution of descriptive variables between non-Hispanic white and Hispanic women, as well as between Hispanic subgroups, included t-test, chi-square test, and Wald’s F-test.
Multiple linear regression models were then estimated for both the five-year and lifetime absolute breast cancer risk estimates, separately, controlling for the explanatory variables. For each risk estimate, two different comparisons were made: 1) between Hispanic vs. non-Hispanic Whites and 2) among Hispanic subgroups. A three step multiple imputation method, previously used in analysis of a US-based population health survey and described22, was used to impute poverty level since about 24% of the observations were missing these values. First, we estimated an ordinal logistic regression model for poverty level using marital status, education and Hispanic subgroup combined with country of birth/time in the US. Second, for each individual with missing poverty level data, we generated probability cut-points for each category of poverty level, based on regression model coefficients. Third, we drew a random number between zero and one from a uniform distribution and compared it to the probability cut-points in order to assign each individual to one category. Noteworthy, the National Center for Health Statistics (NCHS) also generated imputed income variables for both the 2000 and 2005 NHIS surveys23; although, we used the previously described method for this study.
Additionally, a stepwise approach to the regression models was employed, using four different models. Model 1 adjusted for marital status, age, education, federal poverty level; Model 2 adjusted for model 1 covariates plus usual source of care and insurance; Model 3 adjusted for model 2 covariates plus BMI; and Model 4 adjusted for model 3 covariates plus country of origin and years in the US. For the acculturation, a composite variable was generated that combined county of birth (US born, foreign born) and years in the US (<5 years, 5-9 years, ≥10 years). Bivariate analyses indicated that language most often spoken was not predictive of breast cancer risk and, accordingly, was dropped from the final regression analyses. All computations were conducted using SAS software version 9.2 (SAS Institute Inc, Cary, North Carolina) and SAS callable SUDAAN version 9.0 (RTI, Research Triangle Park, North Carolina).
Non-Hispanic white women had a significantly higher mean age compared to Hispanic women (54.3 years vs. 50.3 years, respectively, p<0.001) (Table 1). Overall, a significantly greater proportion of Hispanics were younger than age 12 years at initiation of menarche, younger at first live birth, had no family history of breast cancer, and had never received a breast biopsy compared to non-Hispanic white women. Hispanic women, overall, had a significantly higher proportion with less than a high school education (46.0%) and had a household income that was less than 100% of the federal poverty level (21.7%) compared to non-Hispanic whites (11.9% and 6.8%, respectively).
Among Hispanic subgroups, Cuban/Cuban-American women had the highest mean age, 56.5 years. A greater proportion of Mexican/Mexican-American (48.3%) and Other Hispanic women (86.5%) were born in the US, while Puerto Rican women had the greatest proportion that spoke only/mostly Spanish (27.4%). Puerto Rican (91.4%) and Other Hispanic (93.8%) women had the greatest proportion with a usual source of care, and Central/South American had the greatest proportion of uninsured women (52.8%). Mexican/Mexican-American (36.4%) and Puerto Ricans (36.7%) had the greatest proportion of obese women among Hispanic subgroups.
Table 2 presents the mean 5-year and lifetime BCRAT absolute risk estimates of study participants. These results indicate that, for Hispanic women in our study, on average, the probability of developing invasive breast cancer over the next five years was 0.64% and over their lifetime was 5.88%, which is significantly lower than for non-Hispanic whites (1.24% and 8.63%, respectively, p<0.001). Among Hispanic women, Cuban/Cuban-American and Other Hispanic women had a significantly higher mean BCRAT 5-year absolute risk compared to Mexican/Mexican-American women (p<0.001). Dominican and Central/South American women had a significantly higher mean BCRAT lifetime absolute risk compared to Mexican/Mexican-American women (p<0.001).
A significantly lower proportion of Hispanic women were at high risk of breast cancer compared to non-Hispanic white women (p<0.001, Table 3), based on both the BCRAT 5-year and lifetime risk estimate (e.g. the threshold values for identifying women at high-risk, as defined by ASCO and NCCN, of ≥1.67% 5-year risk of invasive breast cancer and ≥20% lifetime risk of invasive breast cancer). Approximately 2.6% of all Hispanic women had BCRAT 5-year absolute risk ≥1.67% and only 0.2% had a lifetime absolute risk ≥20.0%. Central/South American women had a significantly lower proportion of women at high risk of breast cancer (1.0%) compared to Mexican/Mexican-American women (2.7%, p<0.001). Among Hispanic subgroups, few women had a BCRAT lifetime absolute risk ≥20.0%, with only 0.4% of Other Hispanic women and 0.3% of Mexican/Mexican-American women meeting this high-risk threshold.
Table 4 summarizes the results of the multivariate regression models, indicating that Hispanic women had significantly lower 5-year and lifetime absolute risk of developing breast cancer, based on the BCRAT, compared to non-Hispanic whites (p<0.001). Hispanic women’s mean 5-year absolute risk of invasive breast cancer was consistently lower than for non-Hispanic white women (by an difference of 0.4 percentage points; p<0.001), while the mean lifetime absolute risk of invasive breast cancer for Hispanics was approximately 2.8 percentage points lower than for non-Hispanic white women (p<0.001).
In regard to Hispanics, Cuban/Cuban-American women’s mean 5-year absolute risk of breast cancer was higher than the risk for Mexican/Mexican-American women (p<0.05). No other differences in BCRAT 5-year absolute risk were observed among Hispanic subgroups, when compared to Mexican/Mexican-American women. In assessing lifetime absolute risk of breast cancer, Dominican women’s mean risk was significantly higher than the mean risk for Mexican/Mexican-American women, and remained significant after controlling for all covariates (p<0.001). There were no other differences observed in lifetime absolute risk when comparing other Hispanic subgroups to Mexican/Mexican-American women.
Comparing all Hispanics to non-Hispanic whites, age, education, poverty level, insurance status and BMI were all significantly associated with women’s 5-year absolute risk of breast cancer (results not shown). In particular, increased age, having public health insurance and being overweight/obese were all associated with significant increases in 5-year absolute risk. Similar trends were observed for lifetime absolute risk of breast cancer, although increased age was associated with a significantly lower lifetime absolute risk of breast cancer. For comparisons among Hispanic subgroups, age, marital status, education, poverty level, insurance status and years in the US were significantly associated with 5-year absolute risk of breast cancer (results not shown). Specifically, women of the same Hispanic subgroup who had lived in the US for less than 5 years, had a significantly lower 5-year absolute risk of breast cancer compared to those women born in the US (p<0.05), controlling for all other covariates. This association was not observed for lifetime risk of breast cancer.
Building on current literature, the present study was a broad assessment of the absolute risk of developing breast cancer among different subgroups of Hispanic women. Our findings indicate that, overall, Hispanic women’s 5-year and lifetime absolute risk of developing invasive breast cancer, based on the BCRAT, were significantly lower compared to non-Hispanic white women. Among Hispanic subgroups, Cuban/Cuban-American women had a greater BCRAT 5-year absolute risk of developing invasive breast cancer, while Dominican women had a greater BCRAT lifetime absolute risk of invasive breast cancer. Factors including age, education, health insurance, usual source of care, poverty level, and length of residence in the US were associated with differences in breast cancer risk among Hispanic subgroups. Furthermore, the proportion of Hispanic women who may be eligible for breast cancer risk reduction strategies such as prophylactic tamoxifen and raloxifene use, based on the American Society of Clinical Oncology (ASCO) and National Comprehensive Cancer Network (NCCN) guidelines, ranges from about 1% of Central/South American women to 4% of Puerto Rican women.
Consistent with previous findings24, we found that Hispanic women have a significantly lower 5-year and lifetime absolute risk of developing invasive breast cancer, based on the BCRAT. Further, the magnitude by which Hispanic women’s breast cancer risk was lower than non-Hispanic whites remained significant, and relatively constant, even after multivariate adjustment for key risk factors. These findings suggest other factors, either unidentified and/or unmeasured in this analysis, may help explain the differences observed in BCRAT risk estimates between Hispanic and non-Hispanic white women.
Evidence on the risk of developing breast cancer risk among Hispanics has highlighted differences between US and foreign-born women15, with results from the San Francisco Bay Area Women’s Breast Cancer Study showing that foreign-born Hispanic post-menopausal women had a significantly lower risk of breast cancer compared to their US-born counterparts. Our study expands on this by evaluating differences in risk by Hispanic subgroup, finding that, compared to Mexican/Mexican-American women, Cuban/Cuban-Americans had a significantly higher 5-year risk of breast cancer, whereas Dominicans had a significantly higher lifetime risk of breast cancer. Consequently, while BCRAT risk among other Hispanic subgroups did not differ, the increased risk among Cuban/Cuban-American and Dominican women may reflect other important underlying factors, such as genetic ancestry 25, that account for differences in BCRAT estimates of the risk of developing invasive breast cancer among these Hispanic subgroups.
Our finding of lower breast cancer risk in foreign-born Hispanic women, with less than five years of residence in the US, further strengthens the importance of birthplace, migration and length of residence in the US as factors associated with breast cancer risk among Hispanic women15, 26. Therefore, migration-related changes experienced by Hispanic women who move to the US, such as those to hormonal and lifestyle factors (i.e. weight gain or changes in diet), may have important ramifications on the risk of developing breast cancer.
Noteworthy, our study is among the first, if not only, to provide an estimate of the proportion and number of women from different Hispanic subgroups who may be eligible for prophylactic tamoxifen and raloxifene use. Both ASCO and NCCN recommend that patients who meet the high risk threshold of BCRAT 5-year risk ≥1.67% are eligible for, and may consider, counseling about propylactic tamoxifen or raloxifene to reduce the risk of developing invasive breast cancer in the future27, 28. Thus, as indicated by our findings, up to 2.7% of Mexican-American women (approximately 99,000 women) or 3.7% of Puerto Rican women (approximately 30,000 women), among others, may benefit from risk-reduction counseling to consider their options for preventing the onset of breast cancer.
In interpreting the results of this study it is important to acknowledge its strengths and limitations. The NHIS is a large, population-based study, which allowed us to explore US Hispanic women; however, our study was limited by small samples of women from certain Hispanic subgroups. Nevertheless, these findings build upon those studies of breast cancer risk among different subgroups of Hispanic women, adding considerably to the sparse literature documenting the heterogeneity of the Hispanic population. A possible limitation of this study is that the NHIS is a cross-sectional study and, therefore, we are not able to draw any causal inference on the relationship between breast cancer risk and the risk factors included in our analysis. Another limitation is the use of the NCI BCRAT to estimate breast cancer risk, which has been shown to underestimate risk among US Hispanic women29. Despite such underestimation, our findings represent a conservative estimate of breast cancer risk among Hispanic women and the number of women in each subgroup who could benefit from risk-reduction strategies.
In summary, our findings indicate that, based on the BCRAT, Hispanic women have a significantly lower absolute risk of developing invasive breast cancer, compared to non-Hispanic white women; furthermore, we highlight differences in the BCRAT risk among women in different Hispanic subgroups. While country of origin and length of residence were significantly associated with BCRAT risk estimates in our study sample, it is imperative to further investigate how Hispanic ancestry and migration impact women’s risk factors for developing breast cancer. Lastly, we provide national estimates of the number of Hispanic women, from six key subgroups, who would be eligible for counseling to consider risk reduction by prophylactic breast cancer chemoprevention, as recommended by ASCO and NCCN. Future studies are warranted that further investigate the impact of breast cancer among Hispanics of different ancestry, as well as other important breast cancer risk factors, such as history of cancer in second and higher-degree relatives in this population.
The authors would like to thank Dr. Mitchell Gail and Dr. Jeremy Steeves for their input on the manuscript. M.P.Banegas conducted this work while at the University of Washington and Fred Hutchinson Cancer Research Center and was supported, in part, by the National Cancer Institute Biobehavioral Cancer Prevention and Control Training Program (R25CA092408) at the University of Washington and the National Cancer Institute Center for Hispanic Health Promotion Training Program (1U54CA153502-01) at the Fred Hutchinson Cancer Research Center.
Competing Interests: None declared.