Background: Adult body size has long been known to influence breast cancer risk, and there is now increasing evidence that childhood and adolescent body size may also play a role.
Methods: We assessed the association with body size at ages 10, 15, and 20 years in 475 premenopausal and 775 postmenopausal Hispanic women who participated in a population-based case–control study of breast cancer conducted from 1995 to 2004 in the San Francisco Bay Area. We used unconditional logistic regression to estimate ORs and 95% CIs for the associations with self-reported relative weight compared with peers and body build at ages 10, 15, and 20 years.
Results: In premenopausal women, we found inverse associations with relative weight compared with peers, with ORs of 0.63 (Ptrend = 0.05), 0.31 (Ptrend < 0.01), and 0.44 (Ptrend = 0.02) for heavier versus lighter weight at ages 10, 15, and 20 years, respectively. These inverse associations were stronger in currently overweight women and U.S.-born women and did not differ significantly for case groups defined by estrogen receptor status. In postmenopausal women, not currently using hormone therapy, inverse associations with relative weight were limited to U.S.-born Hispanics.
Conclusions: Large body size at a young age may have a long-lasting influence on breast cancer risk in premenopausal, and possibly postmenopausal, Hispanic women that is independent of current body mass index.
Impact: These findings need to be weighed against adverse health effects associated with early-life obesity. Cancer Epidemiol Biomarkers Prev; 20(12); 2572–82. ©2011 AACR.
Studying migration patterns and breast cancer risk in Hispanic women, we have previously reported a trend of increasing risk with earlier age at migration to the United States (1), suggesting that migration-related changes in early-life exposures contribute to the increasing breast cancer incidence rates in Hispanic immigrants. It has been shown that body size is one factor that can change substantially with migration, particularly in children and adolescents (2).
Adult body size has long been shown to be an important modifiable risk factor for breast cancer, with dual effects depending on menopausal status. Large body size in adulthood is inversely associated with breast cancer risk in premenopausal women (3), whereas in postmenopausal women not using hormone therapy (HT), body size is positively associated with risk (4). More recent epidemiologic studies also suggest a possible etiologic role for body size during adolescence, when developing mammary glands may be particularly susceptible to breast carcinogenesis (5–7). Tall height and rapid change in height in childhood and adolescence have been associated with increased breast cancer risk later in life (8, 9). In contrast, high adolescent body mass index (BMI; refs. 9–11), higher relative weight compared with peers (12, 13), and heavy body build in childhood or adolescence (8, 11, 14–16) have been associated with decreased risk, although results are not consistent across studies (17, 18). It remains uncertain whether the relation between breast cancer and adolescent body size are independent of adult adiposity (11, 14, 19), and whether these associations differ by hormone receptor status (10, 13, 16, 20, 21).
Most prior reports on adolescent body size and breast cancer risk included non-Hispanic white women, with only limited data available for Hispanic women (20). Analyzing data from the San Francisco Bay Area Breast Cancer Study, a population-based case–control study conducted in a multiethnic population, we assessed the association between self-assessed relative and absolute body size at ages 10, 15, and 20 years and breast cancer risk in premenopausal and postmenopausal Hispanic women.
Materials and Methods
This population-based case–control study of breast cancer has been described elsewhere (1, 22). We report here on a subset of cases diagnosed from May 1, 1998 to April 30, 2002 and frequency-matched controls ascertained between 1998 and 2003 for whom information on adolescent body size measures was collected. Briefly, 9,978 newly diagnosed breast cancer cases aged 35 to 79 years were identified through the Greater Bay Area Cancer Registry. Of 8,756 cases who could be contacted (alive, with a valid address and without physician-reported contraindications), 7,704 (88%) completed a brief telephone screening interview that assessed study eligibility and self-reported race/ethnicity. A total of 1,031 cases were selected to participate in the study (all Hispanics diagnosed from 1998 to 2002, all African Americans diagnosed from 1998 to 1999 and a 10% random sample of eligible non-Hispanic whites diagnosed from 1998 to 1999). Of these cases, 931 (90%) completed an in-person interview, including 650 Hispanics, 134 African Americans, and 147 non-Hispanic whites.
Population controls aged 35 to 79 years were identified through random-digit dialing (22). Of 87,035 randomly generated telephone numbers, a household enumeration was obtained for 32,802 (86%) of 38,015 residential phone numbers. From the pool of potentially eligible females, 1,478 controls were randomly selected by frequency matching on race/ethnicity and the expected 5-year age distribution of cases. Of the 1,386 selected controls who could be contacted (alive and with valid address), 1,269 (92%) completed the telephone screening interview that assessed self-identified race/ethnicity and history of breast cancer. Of 1,198 eligible controls without a history of breast cancer, 1,050 (88%) completed the in-person interview, including 766 Hispanics, 137 African Americans, and 147 non-Hispanic whites. Given the relatively small number of African American and non-Hispanic white women with early-life body size information, we restricted the present analysis to Hispanic women.
Trained bilingual, bicultural professional interviewers administered a structured questionnaire in English or Spanish at the home of the participants and collected information on known and suspected breast cancer risk factors. In addition to information on body size (described below), lifetime histories of physical activity from multiple sources (22), usual dietary intake including alcohol consumption (23), and menstrual, reproductive, medical, and breast cancer family histories were assessed. Information on estrogen receptor (ER) and progesterone receptor (PR) status was available from cancer registry records for 86% of cases. 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).
Body size variables
Body build at ages 10, 15, and 20 years was assessed using 9 line drawings of body figures ranging from slim to heavy (24) and categorized as slim (figures 1–2), average (figures 3–4), or heavy (figures 5–9). Weight comparisons with peers at ages 10, 15, and 20 years were assessed using a 5-point scale (1 = much lighter, 2 = lighter, 3 = same, 4 = heavier, and 5 = much heavier) and categorized as lighter (1–2), same (3), or heavier (4–5). Current BMI was calculated as weight (in kg) divided by height squared (in m), based on measured height at interview (or self-reported adult height for 3% of cases and 2% of controls who declined the measurements) and self-reported weight in the reference year (or measured weight for 1% of cases and 4% of controls without self-report). Among controls with both self-reported weight in the reference year and measured weight at interview, the 2 weights were highly correlated (r = 0.91), justifying the substitution of measured weight for self-reported weight when the latter was missing. Adult BMI was classified as normal weight (18.5–24.9 kg/m2), overweight (25.0–29.9 kg/m2), or obese (≥30.0 kg/m2; ref. 25). Women who were underweight (BMI <18.5 kg/m2; 2 cases and 3 controls) were grouped with normal weight women.
We carried out separate analyses for premenopausal and postmenopausal Hispanic women. Premenopausal women included those who were pregnant, breast-feeding, or reported that they were still having menstrual periods in the reference year. The reference year was defined as the calendar year prior to diagnosis (cases) or selection into the study (controls). Women were considered postmenopausal if they reported that their periods had stopped more than 1 year prior to diagnosis (cases) or selection into the study (controls), if they had had a bilateral oophorectomy, or if they were aged 55 years or older at the time of diagnosis or selection and had either started HT use prior to cessation of menses or had had a simple hysterectomy (without oophorectomy). Women for whom menopausal status could not be determined based on these criteria (59 cases, 62 controls) and women with missing information on HT use (8 cases, 2 controls) were excluded from all analyses.
Unconditional logistic regression was used to estimate breast cancer risk associated with adolescent body size measures comparing cases with controls. Polytomous logistic regression was used to compare cases with ER+ or ER− tumors to controls. Analyses by joint ER and PR status (data not shown) were unstable due to small numbers and did not alter our interpretation of the data.
Multivariate adjusted ORs and 95% CIs were estimated, adjusting for age (years, continuous) and other established risk factors for breast cancer, including country of birth (U.S.-born, foreign-born), education (some high school or less, high school graduate or technical school, some college, college graduate), family history of breast cancer in first-degree relatives (no, yes), prior biopsy-confirmed history of benign breast disease (no, yes), number of full-term pregnancies (nulliparous, 1, 2, 3, and ≥4), lifetime breast-feeding (nulliparous, 0, ≤6, 7–12, 13–24, and ≥25 months), age at first full-term pregnancy (nulliparous, ≤19, 20–24, 25–29, and ≥30 years), use of oral contraceptives (0, <5, and ≥5 years), adult height (cm, quartiles among controls), alcohol consumption during the reference year (0, 0.1–4.9, 5–9.9, 10–19.9, and ≥20 g/d), and caloric intake during the reference year (kcal/d, quartiles among controls). Because age at menarche could be an intermediate variable that is affected by early-life body size, we carried out secondary analyses with and without adjustment for age at menarche. Separate analyses were also done adjusting for and then stratifying by current BMI (<25.0, 25.0–29.9, and ≥30.0 kg/m2). In postmenopausal women, we stratified the analyses by HT use (current use vs. past or never use) given evidence of potential heterogeneity in associations with adult BMI according to current HT use (4). Linear trends were assessed across ordinal values of categorical variables. Differences in ORs between the two case groups defined by hormone receptor status were tested using the Wald statistic P value, which was calculated from the polytomous regression model. All tests of significance were 2-sided and P values <0.05 were considered statistically significant. Analyses were done using SAS Version 9.1 (SAS Institute).
For the multivariate analyses, we excluded 20 women (7 cases, 13 controls) with missing information on covariates and 15 women (12 cases, 3 controls) with a daily caloric intake that was considered unreliable (<600 Kcal or >5,000 Kcal), leaving 1,250 women in the analysis (564 cases, 686 controls). Of these, 475 were premenopausal (210 cases, 265 controls), 580 were postmenopausal with past or never use of HT (205 cases, 375 controls), and 195 were postmenopausal with current HT use (149 cases, 46 controls).
Characteristics of cases and controls by menopausal status are shown in Table 1. Compared with controls, cases were more likely to have been born in the United States, had a higher level of education, were more likely to report a family history of breast cancer in first-degree relatives or a personal history of benign breast disease, had a younger age at menarche, were more likely to be nulliparous, had fewer full-term pregnancies, a later age at first full-term pregnancy, a shorter duration of breastfeeding, and higher use of oral contraceptives. Cases were taller, less likely to be currently obese, and had higher alcohol consumption and caloric intake. More postmenopausal cases than controls were current HT users. A higher percentage of U.S.-born than foreign-born control women reported heavier relative weight compared with peers at ages 10 years (24% vs. 16%) and 15 years (27% vs. 22%) and heavy body build at age 20 years (33% vs. 24%).
Among premenopausal women, breast cancer risk was inversely associated with relative weight compared with peers, with ORs of 0.63 (Ptrend = 0.05), 0.31 (Ptrend < 0.01), and 0.44 (Ptrend = 0.02) for heavier vs. lighter relative weight at ages 10, 15, and 20 years, respectively (Table 2). These associations were independent of current BMI. Additional adjustment for age at menarche did not alter the results (data not shown). Inverse associations with relative weight were somewhat stronger for U.S.-born than foreign-born premenopausal women, even after adjustment for current BMI, with ORs ranging from 0.22 to 0.26 for heavier versus lighter relative weight at ages 10, 15, and 20 years (Table 3). Among foreign-born premenopausal women, relative weight at ages 15 and 20 years were also inversely associated with risk.
Considering the joint effects of current BMI and adolescent body size (Table 4), we found that among overweight women (BMI ≥25 kg/m2), breast cancer risk decreased with heavier relative weight at ages 15 (Ptrend < 0.01) and 20 years (Ptrend = 0.02) and heavy body build at age 20 years (Ptrend = 0.02). There were very few currently normal-weight women (BMI <25 kg/m2) who reported having a heavy body build or heavier relative weight compared with their peers at ages 10, 15, or 20 years; comparisons in this currently normal-weight group revealed no significant associations with adolescent body size. Compared with currently normal-weight (BMI <25 kg/m2) women who were lighter than their peers at age 10 years, risk was lowest in currently overweight women (BMI ≥25 kg/m2) who were heavier than their peers at age 10 years (OR = 0.31, 95% CI: 0.14–0.68), with similar corresponding ORs for relative weight at ages 15 and 20 years (OR = 0.23, 95% CI: 0.10–0.51, and OR = 0.25, 95% CI: 0.12–0.52, respectively).
When stratifying the analyses by ER status (Table 4), premenopausal cases with ER+ disease were more likely to report a lighter relative weight and slim body build at ages 10, 15, and 20 years, with significant inverse associations found for relative weight at ages 10 (Ptrend = 0.04) and 15 years (Ptrend < 0.01). Similarly, although numbers were limited for premenopausal cases with ER− disease, we found that these cases were more likely to report a lighter relative weight at ages 10, 15, and 20 years or a slim body build at age 20 years. Overall, inverse associations did not differ significantly for ER+ and ER− disease (Pheterogeneity > 0.05 for all comparisons).
In postmenopausal women, there were no significant inverse associations with childhood or adolescent body size measures in women not currently using HT (Table 2), both among never HT users and past HT users (data not shown), or in women currently using HT (data not shown). Among women not currently using HT, inverse associations with large body size measures were limited to U.S.-born Hispanic women only (Supplementary Table S1). We found inverse trends for relative weight at ages 10 (Ptrend = 0.01), 15 (Ptrend = 0.08), and 20 years (Ptrend = 0.04) and body build at ages 10 (Ptrend = 0.08) and 15 years (Ptrend = 0.05) that were slightly attenuated after adjustment for current BMI. No statistically significant associations were found for ER+ or ER− tumors in postmenopausal women (Supplementary Table S1).
Similar to earlier reports in non-Hispanic white women, we found that early-life body size influences risk of premenopausal breast cancer in Hispanic women. There was a more consistent pattern of inverse associations for relative weight than for body build. U.S.-born controls were more likely than foreign-born controls to report large childhood and adolescent body size, and particularly strong inverse associations with relative weight were found in U.S.-born women and women with a current BMI ≥25 kg/m2. We found no associations with childhood or adolescent body size in postmenopausal women, except for some suggestive inverse associations in U.S.-born women not currently using HT.
Consistent with our findings, the only other study that reported on adolescent body size and breast cancer risk in Hispanic women also found a suggestive inverse association with high BMI (≥22.5 vs. < 18.2) at age 15 years (OR = 0.65, 95% 0.39–1.08, Ptrend = 0.09; ref. 20). Our findings for premenopausal breast cancer are also consistent with previous reports in mostly non-Hispanic white women. Inverse associations have been found for high relative weight (12, 19), heavy body build (15, 16), or high BMI (10) at various ages in childhood or adolescence, although some studies did not find associations with adolescent body size (11, 17, 18). We found inverse trends for relative weight at ages 10, 15, and 20 years and, consistent with other studies (15, 16), we found that the inverse association of premenopausal breast cancer risk with early-life body size was independent of current BMI, a strong predictor of premenopausal breast cancer risk in this study (26). Similar to other cohort (19) and case–control (27) studies, we found the greatest reduction in premenopausal breast cancer risk in women who were heavy both in childhood or adolescence and adulthood.
Unlike some other studies that reported reduced risks of postmenopausal breast cancer associated with large adolescent body size (11, 13, 16, 28), this study found little evidence of inverse associations in postmenopausal women, regardless of HT use, except for suggestive inverse associations among U.S.-born women not currently using HT. In the Nurses' Health Study (16), the largest study on adolescent body size and breast cancer risk to date, inverse associations did not differ between pre- and postmenopausal women.
Given the small number of African American and non-Hispanic white subjects for whom we collected information on early-life body size, we restricted the analysis to Hispanic women. When we included women of all 3 race/ethnicities, the results for premenopausal women (276 cases, 345 controls) and postmenopausal women not currently using HT (312 cases, 511 controls) were essentially the same (data not shown).
Few studies have assessed whether the association with adolescent body size is modified by hormone receptor status (10, 13, 16, 20, 21) and findings are inconsistent. In the Nurses' Health Study (16), the inverse association with adolescent body build among pre- and postmenopausal women combined was stronger for ER− tumors than ER+ tumors, and PR status was not found to be an important modifying factor. Other studies have also reported stronger inverse associations for ER− tumors with BMI at age 15 years in Hispanics (pre- and postmenopausal women combined; ref. 20) and with body build at age 7 years in a population of Swedish postmenopausal women (21). In contrast, some studies found inverse associations with high BMI at age 18 years limited to ER+ premenopausal breast cancer (10) and with high relative weight at age 12 years limited to PR− postmenopausal breast cancer (13). In this study, inverse associations of premenopausal breast cancer risk with large childhood and adolescent body size were more consistent for ER+ tumors, but associations were not significantly different by ER status.
Some potential limitations need to be considered when interpreting our results. The analyses for premenopausal women, particularly those for breast cancer defined by ER status, were limited by small sample sizes. Our findings suggest potentially different effects of early-life body size on breast cancer risk in U.S.-born versus foreign-born Hispanic women, but our sample size was too small to examine these complex interactions. Exposure ascertainment required recall of body size many years in the past and, thus, could be subject to errors in recall. To minimize recall error, we asked about comparative weight, which may be easier to recall than absolute weight at young ages (29), and we used visual aids, such as line drawings of body figures, that have been validated as a reliable self-reported measure of adult (29) and adolescent (28, 30) body size. Other studies have shown that self-reported body build at 18 years is strongly correlated with recorded BMI at age 18 years (28) and that BMI measured at menarche is strongly correlated with body figures recalled 30 years later (30). It is reassuring that, in our study, relative weight at age 10 years (among control women who had not started menstruation by age 10 years) was inversely associated with age at menarche, with mean ages of 13.4, 13.0, and 12.7 years for lighter, same, and heavier relative weight, respectively. This inverse trend is consistent with other reports (16, 21). When we adjusted the multivariate analyses for age at menarche, the inverse associations were not altered. Our findings on adolescent body size were stronger for relative weight compared with peers than for body build assessed by figure drawings, possibly due to better recall of the former adolescent body size measure. There is some evidence that birth size and infant growth may be related to breast cancer risk later in life and that the effect of adolescent body size may be mediated through these very early-life factors (31, 32). We did not collect information on birth weight or infant growth and, therefore, could not consider these factors in our analyses.
The biological explanations for the inverse association between large adolescent body size and premenopausal breast cancer risk remain uncertain. Childhood obesity is associated with earlier menarche, a well-established risk factor for breast cancer. This association seems incongruent with the observed inverse association with large adolescent body size. Several underlying mechanisms have been postulated (33). It has been suggested that overweight adolescents experience slower pubertal growth and sexual maturation (8), despite having, on average, an earlier onset of puberty (34). Overweight adolescents may also experience a longer interval between menarche and the onset of regular menstrual cycles (14), which may be indicative of a greater frequency of anovulatory cycles (6) and, thus, lower cumulative exposure to ovarian sex hormones. Obesity in young girls has been associated with lower sex hormone-binding globulin levels (35) and higher androgen levels (6, 35); estrogen and progesterone concentrations, on the other hand, do not seem to vary with BMI in adolescence (35, 36). Obesity in young girls has been associated with higher levels of insulin (37) and insulin-like growth factor-1 (38), and it has been hypothesized that obesity at a young age could influence the metabolic set point for life through an insulin-related pathway (20). Findings on hormone levels in young girls, however, are not consistent and not necessarily compatible with decreased breast cancer risk in adult life, suggesting complex biological pathways. Furthermore, it is not known whether hormonal changes associated with adolescent obesity persist in adult life. It has also been suggested that higher levels of estrogen and other hormones in young overweight girls may induce earlier differentiation of breast tissue, with terminally differentiated cells being less susceptible to malignant transformation (16), and that body fat at a young age influences the histologic constituents of breast tissue (11).
Risk factors for breast cancer in Hispanic women have not been extensively studied. Interestingly, in this analysis, the inverse associations with early-life body size in premenopausal women were stronger for U.S.-born than foreign-born Hispanics, and in postmenopausal women not currently using HT, they were limited to U.S.-born Hispanics. The basis for this difference by country of birth is unclear; it is possible that correlates of and contributors to childhood body size differ across these populations. Large body size in childhood or adolescence was more frequently reported by U.S.-born than foreign-born Hispanic women. Similarly, in a study of children from Mexican descent, the prevalence of obesity was considerably higher in those living in California than in Mexico (2). It is possible that the earlier onset of childhood obesity in U.S.-born Hispanics exerts a greater protective effect on breast cancer risk than in foreign-born Hispanics. U.S.-born and foreign-born Hispanics also differ in other factors; foreign-born controls had higher levels of physical activity before age 20 years, later age at menarche, and shorter height (unpublished data), suggesting complex interactions between several early-life factors.
Given increasing trends of obesity in Hispanic children and adolescents (39), it remains to be seen whether or not childhood and adolescent obesity today has the same underlying pathology or biological consequences as it did in the past and, thus, whether it will lead to lower breast cancer incidence rates in the future. Furthermore, the potential benefit of adolescent obesity in reducing breast cancer risk need to be weighed against obesity-related risks of other chronic diseases, such as diabetes, which is increasingly prevalent in obese children and adolescents (40).
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
This work was supported by the National Cancer Institute (NCI) grants R03 CA121875, R01 CA63446, and R01 CA77305; by the U.S. Department of Defense (DOD) grant DAMD17-96-1-6071; and by the California Breast Cancer Research Program (CBCRP) grant 7PB-0068. AIP was supported by the NCI grant R25 CA94880.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: Supplementary data for this article are available at Cancer Epidemiology, Biomarkers & Prevention Online (http://cebp.aacrjournals.org/).
- Received September 1, 2011.
- Revision received October 5, 2011.
- Accepted October 7, 2011.
- ©2011 American Association for Cancer Research.