In postmenopausal women not currently using HT, weight gain was positively associated with risk of ER+PR+ BC and was a stronger predictor of risk than current BMI. The highest elevations in risks were found in subgroups of women with a low young-adult BMI or ≥15 years since menopause. Young-adult obesity was associated with reduced BC risk. High waist circumference and WHtR were associated with increased BC risk, independent of current BMI. Associations with weight gain and young-adult BMI were stronger for NHWs than Hispanics and AAs, whereas associations with waist and WHtR were present only in Hispanic and AA women.
Consistent with other reports (White et al. 2012; Huang et al. 1997; Ahn et al. 2007; Feigelson et al. 2004), we found that weight gain was an important risk factor for postmenopausal BC, independent of current BMI. For current BMI no association remained after adjustment for weight gain. In agreement with other studies (Vrieling et al. 2010), we found that the relation with weight gain was limited to ER+PR+ BC. Risk was increased two-fold for currently obese women (BMI ≥30 kg/m2) who had a young-adult BMI <22.4 kg/m2, which is in agreement with other studies (Ahn et al. 2007; Canchola et al. 2012). We found modest effect modification by young adult BMI for weight gain, though some other studies did not (Barnes-Josiah et al. 1995; Feigelson et al. 2004; van den Brandt et al. 1997; Lahmann et al. 2005). In contrast, BC risk was not increased in women who were obese throughout their adult life, consistent with other (Ahn et al. 2007; Canchola et al. 2012), but not all reports (Barnes-Josiah et al. 1995). In agreement with other reports (Chu et al. 1991; Magnusson et al. 1998; Macinnis et al. 2004), we found time since menopause to be another important modifying factor, with two-fold increased risks of ER+PR+ BC with weight gain only among women with ≥15 years since menopause. Weight gain during adulthood largely reflects an increase in body fat which serves as an important source of estrogen production in postmenopausal women (Siiteri 1987). The role of an estrogen-related pathway is further supported by the observation that the associations with BMI and weight gain are limited to ER+PR+ tumors (Suzuki et al. 2009; Vrieling et al. 2010).
Prior findings in AA and Hispanic women for BMI are inconsistent. In our study, OR estimates were increased only for ER+PR+ disease and were of similar magnitude in the three racial/ethnic groups. Two studies in AAs reported elevated risks of ER+PR+ disease (Palmer et al. 2007; Berstad et al. 2010), and, similar to our study, there was no evidence of an association with BMI for BC overall. Other studies did not consider hormone receptor status (White et al. 2012; Schatzkin et al. 1987; Adams-Campbell et al. 1996; Hall et al. 2000; Zhu et al. 2005), and not all found a positive association with BMI (White et al. 2012; Schatzkin et al. 1987; Zhu et al. 2005). In black women from Nigeria (Ogundiran et al. 2010; Okobia et al. 2006; Adebamowo et al. 2003) and Barbados (Nemesure et al. 2009), no associations with BMI were found. In Hispanic women, BMI was not associated with BC overall (Wenten et al. 2002; Slattery et al. 2007; White et al. 2012) and ER+ disease (Slattery et al. 2007), even among women not using HT (Slattery et al. 2007; White et al. 2012).
We found that high weight gain was associated with a two-fold increased risk of ER+PR+ BC in NHW women. In AAs and Hispanics, the associations were much weaker, likely due to the higher prevalence of young-adult obesity in these groups. Of three studies in AAs that reported on weight gain and BC risk (White et al. 2012; Zhu et al. 2005; Palmer et al. 2007), only one found a significant association with BC risk overall (White et al. 2012). In Hispanic women from New Mexico, large weight gain was associated with a two-fold increased risk of ER+PR+ disease and, for BC overall, a significant trend with weight gain was limited to women with a BMI <22 kg/m2 at age 18 years (Wenten et al. 2002). Two other studies in Hispanics found no association with weight gain (Slattery et al. 2007; White et al. 2012). In order to address these inconsistent findings with BMI and weight gain for AA and Hispanic women, future studies should examine the modifying effect of young-adult obesity with larger sample sizes. This is particularly important since AA and Hispanic women have a higher prevalence of young-adult obesity than NHWs, as found in our study as well as others (Flegal et al. 2010).
Our finding of a strong inverse association of postmenopausal BC risk with high young-adult BMI, which was independent of weight change or current BMI, is consistent with other reports (White et al. 2012; Palmer et al. 2007; Berstad et al. 2010; Barnes-Josiah et al. 1995; Huang et al. 1997; Morimoto et al. 2002; Ahn et al. 2007; Chu et al. 1991; Brinton & Swanson 1992; Magnusson et al. 1998), although in some studies there was no association with young-adult BMI (Canchola et al. 2012; Feigelson et al. 2004; Lahmann et al. 2005). In agreement with a large meta-analysis (Suzuki et al. 2009), we found that the inverse association with young-adult BMI did not differ by tumor hormone receptor status. We further found an inverse association, regardless of HT use, as reported by others (Morimoto et al. 2002; Ahn et al. 2007). These findings do not support an estrogen-related mechanism underlying the association with young-adult BMI.
Abdominal adiposity has been proposed to be more important in estrogen production than adiposity at other body sites (Pinheiro et al. 2009). Studies in primarily NHW women, however, have produced inconsistent results (World Cancer Research Fund / American Institute for Cancer Research 2007). Not all studies considered HT use or ER/PR status, or adjusted for overall adiposity (Canchola et al. 2012; Potter et al. 1995; Huang et al. 2000). Our findings confirm previous reports of positive associations with waist circumference and WHtR only in women not currently using HT (Morimoto et al. 2002; Huang et al. 1999; Friedenreich et al. 2002) and stronger associations for ER+PR+ disease (Canchola et al. 2012; Potter et al. 1995; Huang et al. 2000). We found no association with WHR, whereas elevated WHtR was associated with increased risk, in agreement with another study (Canchola et al. 2012). Both waist circumference and WHtR may be better measures of abdominal adiposity than WHR (Molarius & Seidell 1998; Rankinen et al. 1999). In some studies, the association with abdominal adiposity was attenuated after adjustment for BMI (Morimoto et al. 2002; Lahmann et al. 2004; Tehard & Clavel-Chapelon 2006), whereas in our study associations became stronger after BMI-adjustment. Among women with BMI <25 kg/m2, large waist circumference and high WHtR were associated with two-fold increased risks of BC overall. Contrary to another study that reported an association between abdominal adiposity and ER+PR+ tumors only in normal-weight women (Canchola et al. 2012), we found elevated ORs for ER+PR+ tumors, regardless of BMI, with statistically significant estimates in overweight/obese women.
Unlike NHW women, for whom we found no associations with waist circumference and WHtR, AAs and Hispanics had two- to three-fold increased risk of ER+PR+ BC. We found no association with WHR in any racial/ethnic group. This latter finding is consistent with two studies in AA women (Hall et al. 2000; Palmer et al. 2007), but contrasts with reports from Nigeria (Ogundiran et al. 2010) and Barbados (Nemesure et al. 2009), where large waist circumference and high WHR increased BC risk. Similarly, the two-fold increased risk of ER+PR+ disease associated with large hip circumference that we observed for NHW and Hispanic women was not seen in AAs, whereas previous studies from Nigeria (Ogundiran et al. 2012) and Barbados (Nemesure et al. 2009) reported inverse associations with hip circumference. In the Nigerian study, associations with abdominal adiposity were stronger or limited to women with a BMI <25 kg/m2 (Ogundiran et al. 2012), consistent with our findings. In the only study that examined abdominal adiposity in postmenopausal Hispanic women, hip circumference and WHR were not associated with BC risk in not recent HT users (Slattery et al. 2007), which differs from our finding. Given these conflicting reports, it remains to be determined whether there are true racial/ethnic differences in the effects of abdominal adiposity on BC risk. Abdominal fat comprises different fat stores, and AAs and NHWs differ in abdominal depot-specific body fat (e.g., visceral vs. subcutaneous adipose tissue) (Katzmarzyk et al. 2010). Whether different fat stores affect BC risk differentially has not been examined. Our results suggest that studies should assess multiple measures of adiposity in racially/ethnically diverse populations.
Our analyses of body size and ER-PR- BC risk were limited by small numbers. Consistent with most other studies (Suzuki et al. 2009; Yang et al. 2011), we found no associations with current BMI and weight gain, although there are some reports of inverse (Berstad et al. 2010; Setiawan et al. 2009) or positive associations with BMI (Ritte et al. 2012) and positive associations with weight gain (Canchola et al. 2012). Unlike other studies (Canchola et al. 2012; Potter et al. 1995), we found a strong inverse association with young-adult BMI and ER-PR- disease. Adjustment for BMI strengthened the positive association between waist (Ptrend = 0.07) and hip (Ptrend = 0.01) circumferences and risk of ER-PR- BC, but, consistent with other studies (Canchola et al. 2012; Potter et al. 1995; Huang et al. 2000), we found no significant association with WHR or WHtR. The association between BMI, weight gain, and abdominal adiposity warrants further examination in studies with larger numbers of ER-PR- cases. This is particularly important since few risk factors have been identified for hormone receptor negative BC (Althuis et al. 2004; Ma et al. 2006), which disproportionately affects AA and Hispanic women (Ray & Polite 2010).
Our findings should be considered in light of some limitations. Due to the concern that weight may be impacted by BC diagnosis and treatment, we relied on self-reported weight during the reference year. Although we cannot exclude the possibility of inaccurately recalled weight, among subjects for whom measured and self-reported weight was available, the correlation between the two measures was high (r = 0.85 for cases, r = 0.92 for controls). For waist and hip circumference we had to rely on measurements taken after diagnosis which may have also resulted in misclassification. Finally, some subgroup analyses were limited by small sample sizes, and analyses of potential modifying factors (young-adult BMI, time since menopause) could not be further stratified by race/ethnicity. Larger studies or analyses of pooled data will be necessary to further explore the role of these modifying factors in Hispanics and AAs.
Our study also has several important strengths, including a population-based design, high participation rates among cases and controls in each racial/ethnic group, assessment of both overall and abdominal adiposity, detailed data on established BC risk factors, and availability of data on tumor hormone receptor status for most cases. The racial/ethnic diversity of the study population allowed us to assess associations with body size in Hispanic and AA women, thus contributing to the relatively sparse data in these two racial/ethnic populations that experience a greater burden of obesity than NHWs.
It has been estimated that as many as one third of new postmenopausal BC cases may be attributable to adult weight gain (Huang et al. 1997). Given that a number of BC risk factors relate to events well before menopause (e.g., age at menarche, age at first birth) or characteristics that cannot easily be modified (e.g., BC family history), observed associations with body size suggest possible approaches to lowering BC risk in older women through weight maintenance, avoidance of further weight gain and abdominal adiposity, or weight loss. However, promoting overweight at a young age, as a means of reducing BC risk after menopause, should not be encouraged, given the many adverse health effects associated with obesity, including other cancers (Calle & Kaaks 2004; Chen et al. 2011).