Associations between Weight in Early Adulthood, Change in Weight, and Breast Cancer Risk in Postmenopausal Women

A systematic review of articles published to 2007, concluded that there is convincing evidence that obesity, as measured by body mass index (BMI), increases the risk of postmenopausal breast cancer (1). Evidence that weight during early adult life and weight gain during adult life increases postmenopausal breast cancer risk is less convincing (1).

Five cohort studies (2–6) examined the association between BMI in early adulthood, as reported at study entry, and postmenopausal breast cancer risk. Three of these report higher BMI in early adulthood was associated with a decreased risk (2, 4, 5) and two reported no association (3, 6). None of these studies investigated the association by tumor characteristics. Evidence from a limited number of studies suggest that adult weight gain is positively associated with postmenopausal breast cancer risk (1–3, 5–8), but few studies have investigated its association by age (2, 3) or tumor hormone receptor status (4, 8).

It is also unclear whether the association between BMI at adulthood and risk of postmenopausal breast cancer varies by age. A pooled analysis of cohort studies suggested a stronger positive association for women over the age of 65 years (9), but more recent cohort studies have shown no difference by age (3, 10). Finally, a meta-analysis of case–control and cohort studies showed that BMI was associated with increased risk of oestrogen (ER) and progesterone (PR) positive tumors, but for other tumor subtypes, there were few cohort studies and/or their results were heterogeneous (11).

We used the prospective Melbourne Collaborative Cohort Study (MCCS) to investigate risk of postmenopausal breast cancer in relation to weight and BMI at 18 to 21 years of age, at study entry, and the change in the 2 measures between these 2 points. We also assessed whether any associations with these measures varied by age, hormonal receptor status, and grade of the tumor.

The MCCS is a prospective cohort study of 41,514 people (24,469 women) aged between 27 and 76 years at study entry (99.3% of whom were aged 40–69 years). Participants were recruited between 1990 and 1994 from the Melbourne metropolitan area. Southern European migrants were over sampled to widen the range of lifestyle and genetic variation (13% of the participants were Italians and 11% Greeks). Subjects were recruited via Electoral Rolls (registration to vote is compulsory for adult citizens in Australia), advertisements, and announcements in local media. Southern European migrants were identified using comprehensive lists of Italian and Greek surnames in phonebooks and Electoral Rolls. Further details of the study have been published (12). The Cancer Council Victoria's Human Research Ethics Committee approved the study protocol.

At study entry, weight and height were measured by trained nurses according to written protocols (13). Weight was measured to 100 g using digital electronic scales and height was measured to 1 mm using a stadiometer. Participants were asked to recall their weight when they were between the ages of 18 and 21 years.

Participants were interviewed at study entry about possible risk factors such as menopausal status, country of birth, level of educational attainment, aspects of lifestyle, diet, and reproductive history. Women were considered to have been postmenopausal if they reported not having had a menstrual period in the last 12 months.

Of the 24,469 women in the MCCS, 15,837 (65%) were classified as postmenopausal at study entry. From these, we excluded 379 women who had had invasive breast cancer before study entry, 335 with missing values for weight or height at study entry or recalled weight at age 18 to 21 years, 372 with missing values for any potential confounder, and 310 with extreme values of total energy intake (<1st percentile or >99th percentile). These exclusions left 14,441 women.

Cases were women diagnosed with invasive adenocarcinoma of the breast (International Classification of Diseases for Oncology, 3rd edition, 10th revision rubric C50.0–C50.9) between study entry and December 31, 2010. They were ascertained by record linkage to the Victorian Cancer Registry (VCR) and to the Australian Cancer Database. Deaths were ascertained by record linkage to the Victorian death records and the National Death Index. Addresses and vital status were determined by record linkage to Electoral Rolls, Victorian death records, the National Death Index, from electronic phone books and from responses to mailed questionnaires and newsletters.

The pathology reports held by the VCR were reviewed and cancers classified according to tumor grade and ER, PR, and HER2 status. We measured ER, PR, and HER2 status for tumors (76%) where archival tissue was available. The original diagnostic tumor slides were retrieved from pathology laboratories and reviewed by a single pathologist (C. McLean) who assessed ER, PR, and HER2 status using immunohistochemistry techniques (14). ER and PR tumors were categorized as positive if ≥1% of the nuclei was stained or if the intensity of staining was weak, moderate, or strong. Tumors were categorized as negative if there was no staining. HER2 tumors were categorized as positive if >10% of the nuclei was stained and the intensity of staining was weak, moderate, or strong. Tumors were categorized as negative if ≤10% of nuclei was stained or if there was no staining intensity. The agreements between the ER, PR, and HER2 status assessed by immunohistochemistry and the records held by the VCR were 93%, 74%, and 70%, respectively (for ER, κ = 0.66, P < 0.0001; for PR, κ = 0.36, P < 0.0001; for HER2, κ = 0.24, P < 0.0001). Given the good agreement between the ER, PR, and HER2 data, when archival tumor tissue was not available, ER, PR, and HER2 status was assigned according to the histopathology reports held at the VCR. Data on ER, PR, and HER2 were combined to obtain the tumor subtypes, classified as follows: luminal A if tumor is ER+ or PR+ and HER2−; luminal B if tumor is ER+ or PR+ and HER2+; HER2+ if tumor is ER− and PR− and HER2+; triple negative if tumor is ER− and PR− and HER2−.

Follow-up began at the date of study entry and ended at diagnosis of breast cancer, diagnosis of cancer of unknown primary, death, dates left Victoria or Australia, or December 31, 2010, whichever came first.

Cox regression, with age as the time axis, was used to estimate HRs for weight and BMI at age 18 to 21 years, at study entry and the change in these measures from age 18 to 21 years to study entry, overall, by attained age during follow-up and by tumor characteristics. Tests based on Schoenfeld residuals showed no evidence of violation of the proportional hazard assumption.

Weight was categorized into quartiles and BMI was grouped according to WHO categories (15). Change in weight and BMI were categorized into 4 categories (loss and tertiles of gain).

The models included the following variables obtained at study entry; country of birth (Australia/New Zealand, United Kingdom/Malta, Italy, Greece), level of education (primary education, some secondary education, completed secondary education, degree or diploma), age at menarche (<12, 12, 13, ≥14 years), parity and lactation (nulliparous, parous and never lactated, parous and lactated), oral contraceptive use (never, ever), hormone replacement therapy use (never, past, current), smoking (never, former, current), alcohol intake (lifetime abstention, former, 1–19 g/day, 20–39 g/day, 40 g/day or more), level of physical activity (none, low, medium, and high; see ref. 16, for further details) and total dietary energy intake (as a continuous variable). HRs for weight were also adjusted for height and HRs for weight at study entry and weight change were additionally adjusted for weight at age 18 to 21 years. HRs for BMI at study entry and BMI change were further adjusted for BMI at age 18 to 21 years. Tests for linear trend were based on continuous variables. Fractional polynomial regression of degree 2 was used to test departure from linearity.

Effect modification by age was assessed by splitting attained age into two bands (≤69, >69 years), chosen according to the median age at diagnosis of breast cancer and fitted interaction terms between the anthropometric measures and age bands. We assessed whether the HR for change in weight varied by BMI at age 18 to 21 years by fitting an interaction between the two variables. To test for heterogeneity in the HRs by HRT status, we fitted an interaction term between weight (and BMI) and HRT use. Heterogeneity in HRs by tumor characteristics was assessed by fitting Cox models using a data duplication method (17).

Statistical analyses were conducted using Stata 12.1 (Stata Corporation). P < 0.05 (two-sided) was considered to be statistically significant.

Table 1 summarizes the characteristics of the 14,441 women at study entry. The mean age was 60 years (range, 39–76 years). Over an average of 16.5 years of follow-up between study entry (1990–1994) and December 31, 2010, we identified 668 incident invasive breast cancer cases. The mean age at diagnosis was 69 years (range, 42–86 years).

ER, PR, or HER2 status was known for 635 (95%) cases; 484 (78%) were ER+, 327 (53%) were PR+, and 160 (25%) were HER2+. Grade was known for 605 (91%) cases, which included 119 (20%) well differentiated, 282 (47%) moderately differentiated, and 204 (34%) poorly differentiated tumors.

Table 2 shows HRs for weight and BMI at 18 to 21 years of age, at study entry, and for the change in these measures between 18 and 21 years and study entry. The HRs for weight and BMI at age 18 to 21 years were close to unity. In contrast, the risk was elevated for women with greater weight and BMI at study entry, and the HRs were unaffected by adjustment for weight and BMI at 18 to 21 years. Changes in weight and BMI since age 18 to 21 years were positively associated with breast cancer risk, with little change in the HRs after adjusting for weight and BMI at age 18 to 21 years. The HR for weight loss was 0.62 [95% confidence interval (CI), 0.43–0.88] relative to a gain of <10.1 kg and the HR for a decrease in BMI was 0.62 (95% CI, 0.43–0.88) relative to a gain of <4 kg/m². There was no evidence that the trends for the measures at study entry or their change from age 18 to 21 years departed from linearity (results not shown).

There was no evidence that the association between weight gain and breast cancer risk was modified by BMI at age 18 to 21 years (test for interaction, P = 0.43).

There were no significant differences in HRs for weight and BMI at age 18 to 21 years by tumor grade, ER or PR status, HER2 status, and tumor subtype (results not shown). Similarly, there were no significant differences in HRs for weight and BMI or for the changes in these measures by tumor grade, HER2 status, and tumor subtype (Supplementary Table S1). (For the analyses by tumor subtypes, HER2+ tumor subtypes were dropped due to the small number of cases.)

Table 3 shows HRs by attained age. The HR for BMI at study entry was higher at older ages than for younger ages. The HRs for weight and change in weight and BMI were also greater at older ages, although the statistical evidence was weak (test for homogeneity, P = 0.10–0.16). Because the associations seemed to be restricted to women at least 70 years old, further analyses were restricted to this age stratum. There were no significant differences in HRs by ER status for weight and BMI at study entry and their change from 18 to 21 years and study entry (Table 4). Greater weight and BMI at study entry and greater increases in weight and BMI were associated with increased risk of PR+ tumors but not PR− tumors (all tests for homogeneity, P < 0.01). HRs for ER+/PR+ tumors were elevated for weight and BMI at study entry and change in weight and BMI from age 18 to 21 years to study entry but not for ER+/PR− and ER−/PR− tumors (all tests for homogeneity, P ≤ 0.02). (ER−/PR+ tumors were dropped from the analyses due to the small number of cases.)

The HRs for weight and BMI at study entry and gain in weight and BMI were slightly higher for women who had never used HRT, but the P values for the interactions were large (P ≥ 0.18). The HR for BMI at study entry for never users of HRT was 1.20 per 5 kg/m² (95% CI, 1.09–1.33) whereas for past or current users it was 1.07 (95% CI, 0.92–1.24), but the P value from the interaction was large (P = 0.18). For weight gain, the HR per 5 kg for women who had never used HRT was 1.07 (95% CI, 1.03–1.11) whereas for past or current users the HR was 1.04 (95% CI, 0.98–1.10).

The following additional analyses were conducted but did not materially change the HRs (results not shown): excluding women born in southern Europe; excluding the first 2 years of follow-up; censoring after 10 years of follow-up; using ER, PR, and HER2 measures from immunohistochemistry only or from the VCR only. Our definition of postmenopausal status might include women whose periods had stopped due to a hysterectomy but who might have intact ovaries and so be premenopausal. However, results were similar when restricting the analyses to women whose periods stopped naturally.

Weight and BMI at age 18 to 21 years were not associated with risk of postmenopausal breast cancer regardless of women's ages or the grade or hormonal receptor status of the tumors. Weight and BMI at study entry (39–76 years) and increases in these measures from age 18 to 21 years to study entry were associated with increased risk of breast cancer for women at older ages (>69 years). These associations were restricted to tumors that were positive for ER and PR.

A strength of our study is the almost complete information on hormonal receptor and HER2 status of the tumor. Other strengths of our study are its cohort design and the complete follow-up of participants diagnosed with breast cancer through nationwide cancer registries. We also measured weight and height at study entry. The main limitations of our study are the possibility of poor recall of weight at age 18 to 21 years and the limited precision to assess heterogeneity by tumor subtype. Poor recall would lead to nondifferential error with the consequent attenuation of the association. However, studies that have investigated the validity of self-reported past weight have shown that it is a reasonably reliable measure of the true value (18–22).

Our finding that BMI at study entry was only associated with breast cancer for women who attained the age of at least 70 years during follow-up is consistent with our earlier report based on 9 years of follow-up (23) and with the results of a pooled analysis of 7 cohort studies (9). More recent publications from the Cancer Prevention Study II (CPS-II; ref. 3) and the European Prospective Investigation into Cancer and Nutrition study (EPIC; ref. 10) reported no effect modification by age, whereas the Women's Health Initiative observational study (WHI; ref. 2) reported increased risk for women aged 50 to 69 years but no association for older ages. These differences might be partly explained by the small number of older cases in the WHI study, the shorter follow-up period (<5 years) of the EPIC and WHI studies, and the self-reported measures in the CPS-II.

BMI in early adulthood is inversely related to premenopausal breast cancer risk (4, 24–26), but findings for postmenopausal women are inconsistent. Consistent with our findings, the CPS-II cohort study (3) also reported a null association, but in both studies few women (<1%) had high BMI (>30 kg/m²). The EPIC study (6) reported an inverse association but it was not statistically significant. The Black Women's Health Study (4) and the Miyagi Cohort Study (5) reported inverse associations between early BMI and postmenopausal breast cancer risk. Results from case–control studies coincided with ours (27–30). Several other studies that adjusted for weight at study entry when investigating the association between young adult weight and postmenopausal breast cancer risk are not discussed here because this analytical approach produces bias, as weight at study entry is at least partly a consequence of weight during early adult life.

Six cohort studies (2–7) investigated postmenopausal breast cancer in relation to change in weight from early adulthood using information on both early and current weight obtained at study entry. Consistent with our findings, 5 reported a direct association (2, 3, 5–7), but the other reported no association (4).

We did not observe any subgroup of tumors defined by hormone receptor status for which weight in early adulthood had any effect; reports from cohort studies are sparse (31, 32). A case–control study (33) also found no association by ER/PR status of the tumor.

Our finding of a positive association between weight after menopause and risk of hormone receptor positive breast cancer is consistent with previous studies (11, 23, 31, 34, 35) and supports the hypothesis that the risk of hormone susceptible breast tumors is increased by aromatization of androgens to oestrogens in adipose tissue (36, 37). Also, in heavier women, plasma sex hormone binding globulin is decreased and, in turn, the amount of bioavailable oestrogen is increased (36, 38). For weight gain, the increased risk for hormone receptor positive breast cancer in older postmenopausal women, similar to adult weight, is consistent with results from a meta-analysis (8).

Unlike the California Teachers Cohort Study (39), we found no evidence that BMI at age 18 to 21 years modified the association between postmenopausal breast cancer and change in weight. The lack of association could be due to lack of precision due to small numbers.

For weight at study entry and weight gain, our results show stronger associations for HRT never users that are consistent with previous studies (2, 3, 6, 35, 40, 41). Our tests for heterogeneity by HRT status, however, were not statistically significant. A meta-analysis of studies published from 1994 to 2007 reported that an increase in BMI of 2 kg/m² was associated with an increased risk of postmenopausal breast cancer of 1.05 (95% CI, 1.03–1.07); the estimate was similar among HRT nonusers, but weaker and not statistically significant in HRT users (41). Postmenopausal women with higher BMI have increased adipose tissue which leads to increased levels of endogenous oestrogen and consequently to increased risk of breast cancer (36, 37); HRT might mask the association between weight and weight change and breast cancer risk because it would increase the levels of circulating oestrogens in both lean and obese women, making the contribution from the conversion of androgens into oestrogens occurring in adipose tissue negligible (2).

We have, thus, provided further evidence supporting the hypotheses that postmenopausal weight is a risk factor for hormone receptor positive breast cancer in later life, and that the protective effect of weight early in adulthood observed for premenopausal breast cancer risk does not persist after menopause. Further studies need to evaluate the impact of controlling or reducing weight in adulthood on breast cancer risk at the population level.

No potential conflicts of interest were disclosed.

Conception and design: D.R. English, J.L. Hopper, G.G. Giles

Development of methodology: R.J. MacInnis, D.R. English, J.L. Hopper, G.G. Giles, L. Baglietto

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.R. English, J.L. Hopper, C.A. McLean, G.G. Giles

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Krishnan, J.K. Bassett, R.J. MacInnis, J.L. Hopper, C.A. McLean, G.G. Giles, L. Baglietto

Writing, review, and/or revision of the manuscript: K. Krishnan, J.K. Bassett, R.J. MacInnis, D.R. English, J.L. Hopper, G.G. Giles, L. Baglietto

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases):

Study supervision: L. Baglietto

This study was made possible by the contribution of many people, including the original investigators, the Program Manager, and the diligent team who recruited the participants and who continue working on follow-up. The authors also thank many thousands of Melbourne residents who continue to participate in the study. Cohort recruitment was funded by VicHealth and The Cancer Council Victoria. This study was funded by grants from the National Health and Medical Research Council (251533, 209057, 504711) and The National Breast Cancer Foundation and was further supported by infrastructure provided by The Cancer Council Victoria.

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.