|Home | About | Journals | Submit | Contact Us | Français|
Large body size has been associated with a reduced risk of premenopausal breast cancer in non-Hispanic white women. Data on other racial/ethnic populations are limited. The authors examined the association between premenopausal breast cancer risk and adult body size in 672 cases and 808 controls aged ≥35 years from a population-based case-control study conducted in 1995–2004 in the San Francisco Bay Area (Hispanics: 375 cases, 483 controls; African Americans: 154 cases, 160 controls; non-Hispanic whites: 143 cases, 165 controls). Multivariate adjusted odds ratios and 95% confidence intervals were calculated using unconditional logistic regression. Height was associated with increased breast cancer risk (highest vs. lowest quartile: odds ratio = 1.77, 95% confidence interval: 1.23, 2.53; Ptrend < 0.01); the association did not vary by hormone receptor status or race/ethnicity. Body mass index (measured as weight (kg) divided by height (m) squared) was inversely associated with risk in all 3 racial/ethnic groups, but only for estrogen receptor– and progesterone receptor–positive tumors (body mass index ≥30 vs. <25: odds ratio = 0.42; 95% confidence interval: 0.29, 0.61). Other body size measures (current weight, body build, adult weight gain, young adult weight and body mass index, waist circumference, and waist-to-height ratio) were similarly inversely associated with risk of estrogen receptor– and progesterone receptor–positive breast cancer but not estrogen receptor– and progesterone receptor–negative disease. Despite racial/ethnic differences in body size, inverse associations were similar across the 3 racial/ethnic groups when stratified by hormone receptor status.
Body size is an important and potentially modifiable risk factor for breast cancer (1–4). Associations have been shown to vary greatly across population groups according to menopausal status, menopausal hormone therapy use, and tumor hormone receptor status. In premenopausal women, high body mass index (BMI) is associated with a decreased risk of breast cancer, although growing evidence suggests that the inverse associations are limited to women with tumors that are estrogen receptor– and progesterone receptor–positive (ER+/PR+) (5–10).
Given that most epidemiologic studies of associations between body size and premenopausal breast cancer risk have been conducted in non-Hispanic white women, it is unclear whether these associations hold among other racial/ethnic populations. Studies conducted in premenopausal African American (11–19) and Hispanic (20–22) women are limited and the results are mixed. Some studies found no association with adult obesity (11, 14, 15, 20–22), whereas others reported positive associations (12, 17). We present here the results for BMI and other body size measures in premenopausal women from the San Francisco Bay Area Breast Cancer Study, which was conducted in Hispanic, African American, and non-Hispanic white women.
The methods of this population-based case-control study have been described elsewhere (23, 24). Briefly, 17,581 women aged 35–79 years with incident invasive breast cancer were identified through the Greater Bay Area Cancer Registry. Eligibility was restricted to women residing in Alameda, Contra Costa, San Mateo, San Francisco, or Santa Clara county, California, at diagnosis. Of 15,573 cases contacted (those who were alive, had a valid address, and had no physician refusal), 89% completed a brief telephone screening interview that assessed study eligibility and self-reported race/ethnicity. A total of 2,571 cases were selected (all Hispanic cases diagnosed from April 1, 1995, to April 30, 2002, all African-American cases diagnosed from April 1, 1995, to April 30, 1999, and a 10% random sample of non-Hispanic white cases diagnosed from April 1, 1995, to April 30, 1999). Of these women, 2,258 (88%) completed an in-person interview, including 1,119 (89%) Hispanics, 543 (87%) African Americans, and 596 non-Hispanic (86%) whites.
Controls aged 35–79 years were identified through random digit dialing (23). Of 161,703 randomly generated telephone numbers, household enumeration (for age, sex, and race/ethnicity of each household member) was obtained for 61,576 (84%) of 73,380 phone numbers known to be residential. From the pool of potentially eligible women, 3,771 controls were randomly selected and frequency-matched based on race/ethnicity and the expected 5-year age distribution of cases. Of the 3,547 controls contacted (alive and with a valid address), 92% completed a telephone screening interview that assessed study eligibility (no history of breast cancer, self-identified race/ethnicity). Of 3,170 controls selected, 2,706 (85%) completed the in-person interview, including 1,462 (88%) Hispanics, 598 (82%) African Americans, and 646 (83%) non-Hispanic whites.
The analysis was restricted to premenopausal women. We excluded women who reported natural (805 cases, 995 controls) or surgical (337 cases, 366 controls) menopause, as well as women using hormone therapy (414 cases, 499 controls). Thus, 702 cases and 846 controls were included in the present analysis. All study participants provided written informed consent, and the study was approved by the institutional review board of the Cancer Prevention Institute of California (formerly the Northern California Cancer Center).
Professional interviewers administered a structured questionnaire in English or Spanish at the participants’ homes and collected information on breast cancer risk factors and body size, including adult height, weight in the reference year (defined as the calendar year before diagnosis for cases or before selection into the study for controls), and body build in the reference year relative to 9 figure drawings ranging from slim to heavy (25). For cases diagnosed from April 1995 to April 1998 and their matched controls, information was collected on average weight at ages 25–30 years. For cases diagnosed after April 1998 and their matched controls, information was collected on average weight at ages 20–29 years. The interviewers also took 3 repeated measurements of standing height, hip circumference, and waist circumference and 2 measurements of weight. Height was measured to the nearest millimeter using a stadiometer after study participants removed their shoes. Weight was measured to the nearest 0.2 kg using a portable scale, with study participants wearing light clothing. Waist and hip circumferences were measured to the nearest millimeter using a linen tape measure. Waist circumference was measured at the natural indentation of the waist; hip circumference was measured at the greatest protrusion of the buttocks.
The questionnaire also asked about lifetime history of physical activity, including sports and exercise, walking for transportation, strenuous indoor and outdoor chores, and occupational activity (23). For each activity, we assessed time spent per week and estimated average lifetime physical activity by multiplying the average annual hours per week by the number of years the woman engaged in that activity, summed those quantities across the period from menarche to the reference year, and then divided by the number of years from menarche to the reference year. A food frequency questionnaire adapted from the 106-item Block Health History and Habits Questionnaire, developed in 1995 (26, 27), was used to assess usual dietary intake and alcohol consumption during the reference year. Information on ER and PR status was available from cancer registry records for 85% and 84% of cases, respectively.
Current BMI was calculated as weight (in kilograms) divided by height (in meters) squared. Calculations were based on measured height at interview (or self-reported height for 6% of cases and 5% of controls who declined measurement) and self-reported weight in the reference year (or measured weight for 1% of cases and 2% of controls without self-reported data). BMI in a woman's 20s (young-adult BMI) was based on reported average weight at ages 25–30 years (410 cases, 492 controls) or at ages 20–29 years (292 cases, 354 controls) (as described above) and height measured at interview (or self-reported height, if the measurement was declined). Current body build was assessed using 9 line drawings of bodies ranging from slim to heavy (25, 28). Waist-to-hip ratio (WHR), calculated as measured waist circumference (in centimeters) divided by hip circumference (in centimeters), is a measure of body fat distribution that reflects both adipose tissue and muscle mass; waist-to-height ratio (WHtR), calculated as waist circumference divided by height, measures visceral adiposity independent of height and therefore more directly reflects abdominal adiposity alone (29). Adult weight gain (in kilograms) was calculated as the difference between self-reported weight in a woman's 20s and self-reported weight in the reference year. BMI in all analyses was classified using World Health Organization–defined cutpoints (underweight to normal weight: <25.0; overweight: 25.0–29.9; obese: ≥30.0) (30). For analyses involving all cases, we categorized the body size variables according to the quartile distribution among controls. For race/ethnicity-specific analyses involving ER+/PR+ and ER−/PR− cases only, we categorized the body size variables according to the tertile distribution among controls.
Unconditional logistic regression was used to calculate odds ratios and 95% confidence intervals comparing cases with controls, both overall and separately for each racial/ethnic group. Polytomous logistic regression was used to compare ER+/PR+ (n = 319) and ER−/PR− (n = 171) case groups with controls. Other case groups (ER+/PR− and ER−/PR+) were too small for separate analyses (n = 65 and n = 32, respectively).
Odds ratios in all analyses were adjusted for age (continuous) and the following variables, which were significantly associated with breast cancer risk in our study: country of birth (United States or other), education level (some high school or less, high school graduate or vocational/technical school, some college, or college graduate), first-degree family history of breast cancer (yes or no), biopsy-confirmed history of benign breast disease (yes or no), age at menarche (≤11, 12, 13, or ≥14 years), parity (0, 1, 2, 3, or ≥4), duration of breastfeeding (nulliparous or 0, ≤6, 7–12, 13–24, or ≥25 months), alcohol consumption during the reference year (0, 0.1–4.9, 5–9.9, 10–19.9, or ≥20 g/day), average lifetime physical activity (hours per week, quartiles), daily caloric intake (quartiles), and height (quartiles). Analyses of all women combined were also adjusted for race/ethnicity. Linear trends were assessed across ordinal values of categorical variables. Significant differences in odds ratios between case groups were tested using the Wald statistic P value, calculated from the polytomous regression model. Two-sided P values were reported for tests of trend and tests of heterogeneity, with P values <0.05 considered statistically significant.
The final analysis was based on 672 premenopausal cases and 808 premenopausal controls after excluding 16 women (9 cases, 7 controls) with missing information on covariates and 52 women (21 cases, 31 controls) with a daily caloric intake that was considered unreliable (<600 kcal or >5,000 kcal). Statistical analyses were conducted using SAS version 9.1 software (SAS Institute, Inc., Cary, North Carolina).
Characteristics of premenopausal cases and controls are shown in Table 1. For cases with known hormone receptor status, ER+/PR+ tumors were more common in Hispanics (57%) and non-Hispanic whites (57%) than in African Americans (45%), whereas ER−/PR− tumors were more common in African Americans (37%) and Hispanics (27%) than in non-Hispanic whites (27%). In comparison, among breast cancer cases ≤50 years of age diagnosed in the greater San Francisco Bay area from 1995 to 2002 (the ascertainment period of the case-control study), the distribution by ER/PR status was similar (data not shown); 64% of non-Hispanic white women had ER+/PR+ tumors, compared with 56% of Hispanic women and 41% of African-American women, and 23% of non-Hispanic whites had ER−/PR− tumors, compared with 30% of Hispanic women and 43% of African American women.
Compared with controls, cases were more likely to be US-born and nulliparous and to have a higher education level, a family history of breast cancer, a personal history of benign breast disease, earlier menarche, lower parity, a shorter duration of breastfeeding, a lower lifetime physical activity level, and higher alcohol consumption (Table 1).
Body size characteristics among controls differed considerably by race/ethnicity (Table 2). African Americans and non-Hispanic whites were of similar average height, whereas Hispanic women were, on average, considerably shorter. Young-adult BMIs were similar in African Americans and Hispanics but higher than in non-Hispanic whites. Weight gain of >20 kg was almost twice as frequent in African Americans as in non-Hispanic whites, and the frequency of large current body size (high weight, obesity, and heavy body build) was highest in African-American women. Average WHRs and WHtRs were similar in Hispanics and African Americans and somewhat higher than in non-Hispanic whites. The proportions of women with high waist circumference, WHR, and WHtR were highest in African Americans, intermediate in Hispanics, and distinctly lower in non-Hispanic whites.
Associations between body size and premenopausal breast cancer risk for all women combined and separately by race/ethnicity are shown in Table 3. For all women combined, risk increased with increasing height (Ptrend < 0.01), and strong inverse trends were found for current body size (weight, BMI, and body build), young-adult BMI, and weight gain since young adulthood. Risk reductions were similar across racial/ethnic groups for current weight (highest vs. lowest quartile: odds ratios (ORs) ranging from 0.43 to 0.55), BMI (≥30 vs. <25: ORs ranging from 0.52 to 0.65), and body build (heavy vs. slim: ORs ranging from 0.25 to 0.40). Young-adult BMI was inversely associated with risk in Hispanics (highest vs. lowest tertile: OR= 0.57, 95% confidence interval (CI): 0.39, 0.85; Ptrend = 0.01). A nonsignificant inverse association was also observed in African-American women (OR = 0.64, 95% CI: 0.34, 1.21; Ptrend = 0.17) but not in non-Hispanic white women. Weight gain was inversely associated with risk among all women combined (Ptrend < 0.01); when stratified by race/ethnicity, inverse associations were seen in all 3 racial/ethnic groups, although the trend was significant in Hispanic women only. Weight gain of >20 kg was associated with odds ratios of 0.35 (95% CI: 0.21, 0.59) in Hispanics, 0.52 (95% CI: 0.22, 1.22) in African Americans, and 0.57 (95% CI: 0.22, 1.45) in non-Hispanic whites. We found no association between premenopausal breast cancer risk and waist or hip circumference, WHT, or WHtR in all women combined or in any racial/ethnic group. None of the differences in odds ratios by race/ethnicity were statistically significant (P > 0.05).
Associations with body size measures for ER+/PR+ and ER−/PR− tumors are shown in Table 4. Associations with height were similar for the 2 case groups. For other body size measures, strong significant inverse trends were found only for cases with ER+/PR+ tumors. Statistically significant heterogeneity by hormone receptor status was found for current weight, current BMI, and weight gain, and heterogeneity of borderline significance was observed for WHtR (P = 0.06), current body build (P = 0.09), and waist circumference (P = 0.09). WHR was not associated with risk in either case group.
For ER+/PR+ tumors, similar inverse trends in Hispanic and African American women were found for current weight, BMI, body build, young-adult BMI, and weight change (Table 5). For non-Hispanic white women, inverse trends of borderline significance were found for current BMI, current body build, and waist circumference. The differences in odds ratios by race/ethnicity were not statistically significant (P > 0.05).
The present analysis in a multiethnic population found positive associations between height and premenopausal breast cancer risk, as well as substantial inverse associations with obesity and other measures of body fatness and abdominal adiposity (current weight, body build, waist circumference, WHtR, BMI in young adulthood, and weight gain). Importantly, inverse associations were limited to ER+/PR+ tumors. For all tumors combined and for ER+/PR+ tumors, the inverse associations were similar for Hispanic, African-American, and non-Hispanic white women.
Compared with the large number of studies that have examined the relation between body size and premenopausal breast cancer risk in non-Hispanic white women (4), data are relatively sparse and inconsistent for Hispanic (20–22) and African-American (11–15, 17–19) women. We found inverse associations of similar magnitude across racial/ethnic groups but our sample size, particularly in African Americans and non-Hispanic whites, was too small to determine whether the race/ethnicity-specific odds ratios we observed were actually statistically different. Our analysis is the first to examine associations between multiple body size measures and breast cancer risk in a single study of Hispanic, African-American, and non-Hispanic white women. The population-based design and similar response rates among eligible cases and controls from all 3 racial/ethnic groups increased the generalizability of our results. Furthermore, we found the inverse associations to be consistent across multiple measures of body size and in agreement with previous reports for non-Hispanic white women (3, 31).
The present results should be interpreted in the context of some study limitations. Given the inherent need to rely on self-reporting for most variables, we cannot exclude the possibility of inaccurate recall, which could have resulted in misclassification of confounders and exposure variables. For the assessment of body size, we relied on anthropometric measurements in addition to self-reported measures. Because of concern that cases may have experienced treatment-related weight gain or disease-related weight loss (32), we calculated current BMI from self-reported weight in the reference year instead of measured weight at interview, except for the small proportion of cases and controls who declined to self-report but allowed actual weight measurement. Inaccurate recall or misreporting of weight in the reference year could have biased the odds ratios for current BMI and weight towards the null. For young-adult measures (weight and BMI) and weight gain, we relied on 2 measures of weight in a woman's 20s. In the early version of the questionnaire, we assessed weight at the ages of 25–30 years; in a later version, we asked about weight in each decade (e.g., at ages 20–29 years). In a sensitivity analysis, we found that each of the young-adult measures was inversely associated with risk, regardless of the weight measure used, both in all women combined and in women with ER+/PR+ tumors.
Consistent with some (33–35) but not all (22) studies in non-Hispanic white women, we found a positive association between height and breast cancer risk in Hispanic women. Results in other populations have been mixed as well. Positive associations with height have previously been reported for African-American (15, 16) and Nigerian (36, 37) women. Other studies, like ours, did not find significant trends with height in black (13, 38) or non-Hispanic white (2, 39–41) women.
Our finding of a 40% lower risk of premenopausal breast cancer in obese women is consistent with other studies in non-Hispanic whites (3, 31). Importantly, and in agreement with other studies (42), the inverse association with BMI was limited to ER+/PR+ tumors in all 3 racial/ethnic groups. Prior studies of BMI in African-American women, black women in Barbados, and Hispanic women have produced mixed results. Palmer et al. (18) and Nemesure et al. (38) reported inverse associations with obesity, although in the latter study, no significant association remained after adjustment for BMI at age 18 years. In our analysis of current BMI, additional adjustment for young-adult BMI changed the odds ratio estimates only minimally (data not shown). Other studies found no evidence of inverse associations between BMI and breast cancer risk in African-American (11, 12, 14, 15, 17), Nigerian (36, 37), or Hispanic (20–22) women.
Data on other body size measures are sparse for African-American and Hispanic women. We found substantial inverse associations for current weight and body build, with similar findings in the 3 racial/ethnic groups. Another study found no inverse association between current body build and breast cancer risk in African-American women (19). Large young-adult body size has been associated with lower risk of breast cancer in non-Hispanic white women (40, 43–45), although findings are not consistent (41, 46). We found significant inverse trends for young-adult BMI, with similar findings in African Americans and Hispanics, but no clear trend in non-Hispanic whites. Some (12, 18, 19) but not all (17, 21) other studies in these populations have also reported inverse associations with BMI at age 18 years. BMI at age 18 years may be stronger predictor of premenopausal breast cancer risk than current BMI (18, 19, 43), but the present results were similar for the 2 BMI measures.
We found a strong inverse trend with weight gain since young adulthood, with similar findings in Hispanic and African-American women, and with associations limited to ER+/PR+ tumors. Results from other studies are inconsistent. Studies in non-Hispanic white women suggest that an inverse association with weight gain may be limited to women who experienced their lowest adult weight after age 21 years (40), women with a BMI <20 at age 18 years (19), or women diagnosed with early-stage and lower-grade breast cancer (46). Several studies found no association with weight gain in non-Hispanic white (19, 22, 41, 47, 48), African-American (17, 18), or Hispanic (21, 22) women.
In agreement with other studies (3, 49, 50), we found no association with WHR overall or in any racial/ethnic group. High WHR has been associated with increased risk of premenopausal breast cancer in both African-American and white women from North Carolina (15) and in Nigerian women (37), but most studies, including ours, have found no association (18, 38, 51). Similarly, a study of breast cancer in Hispanic women found no association with WHR (22). However, we found inverse associations with waist circumference and with WHtR for ER+/PR+ tumors.
Overall, our findings are generally consistent across race/ethnicity and are in agreement with previous reports for non-Hispanic white women. Our results also confirm those of some, but not all, previous studies in African-American and Hispanic women. Variability in observed associations across studies may be explained, in part, by differences in the proportion of cases with ER−/PR− tumors, which are more common in African Americans and Hispanics than in non-Hispanic whites. Our data clearly show inverse associations with measures of body size only for ER+/PR+ tumors; no associations were found for ER−/PR− tumors. Similarly, Berstad et al. (19) reported a significant inverse association with BMI in African American women that was limited to ER+/PR+ tumors. No other study in African Americans and Hispanics presented associations with body size measures for premenopausal breast cancer by hormone receptor status. Failure to take hormone receptor status into account may obscure associations with body size and may have contributed to the inconsistent results in African-American and Hispanic women.
The biologic mechanisms underlying associations between large body size and breast cancer risk remain uncertain. The observation that BMI and other body size measures are inversely associated with ER+/PR+ but not ER−/PR− tumors suggests the importance of sex-steroid hormone pathways. Elevated BMI may contribute to lower serum levels of sex hormone–binding globulin (52, 53) and total estradiol (52, 53) and higher levels of free testosterone (53), as well as a higher frequency of anovulatory and irregular menstrual cycles in premenopausal women, which in turn results in reduced production of estrogen and progesterone (54). In 2 recent studies (18, 43), however, the inverse association with large body size was not explained by menstrual cycle characteristics, self-reported infertility, or probable polycystic ovary syndrome, suggesting the importance of other mechanisms.
In conclusion, our results suggest that the inverse association between premenopausal breast cancer and larger body size is similar in Hispanic, African-American, and non-Hispanic white women but is limited to ER+/PR+ tumors. Thus, hormone receptor status is important to consider when evaluating the association between body size and premenopausal breast cancer risk. Although our results indicated a lower risk of premenopausal breast cancer in obese women, repeated studies have noted that weight gain and obesity are associated with increased breast cancer risk during the postmenopausal years, when breast cancer occurs much more commonly than during the premenopausal years; thus, avoidance of weight gain before menopause remains advisable.
Author affiliations: Cancer Prevention Institute of California (formerly the Northern California Cancer Center), Fremont, California (Esther M. John, Meera Sangaramoorthy, Jocelyn Koo, Pamela L. Horn-Ross); Division of Epidemiology, Department of Health Research and Policy, Stanford University School of Medicine, and Stanford Cancer Center, Stanford, California (Esther M. John, Pamela Horn-Ross); and Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, Seattle, Washington (Amanda I. Phipps).
This work was supported by National Cancer Institute grant R03 CA121875. Data collection for the parent study was funded by National Cancer Institute grants R01 CA63446 and R01 CA77305; by US Department of Defense grant DAMD17-96-1-6071; and by California Breast Cancer Research Program grants 4JB-1106 and 7PB-0068. Cancer incidence data used in this publication have been collected by the Cancer Prevention Institute of California (formerly the Northern California Cancer Center) under contract N01-CN-65107 with the National Cancer Institute, National Institutes of Health, and with the support of the California Cancer Registry, a project of the Cancer Surveillance Section, California Department of Health Services, under subcontract 050N-8701/8-S1522 with the Public Health Institute.
The content of this publication does not necessarily reflect the views or policies of the US Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the US government.
Conflict of interest: none declared.