Tel.: 617-732-8496, Fax: 617-732-4899
Adult weight change and incidence of premenopausal breast cancer
Article first published online: 9 MAY 2011
Copyright © 2011 UICC
International Journal of Cancer
Volume 130, Issue 4, pages 902–909, 15 February 2012
How to Cite
Michels, K. B., Terry, K. L., Eliassen, A. H., Hankinson, S. E. and Willett, W. C. (2012), Adult weight change and incidence of premenopausal breast cancer. Int. J. Cancer, 130: 902–909. doi: 10.1002/ijc.26069
- Issue published online: 16 DEC 2011
- Article first published online: 9 MAY 2011
- Accepted manuscript online: 16 MAR 2011 01:46PM EST
- Manuscript Accepted: 7 MAR 2011
- Manuscript Received: 10 DEC 2010
- National Cancer Institute, National Institutes of Health (Public Health Service). Grant Numbers: R01 CA114326, R01 CA50385, R01 CA87969, R25 CA098566
- breast cancer;
- weight gain;
- weight loss;
Overweight and obesity are inversely related to the risk of breast cancer among premenopausal women. We assessed the association between adult weight change since age 18 years with the risk of breast cancer among premenopausal women to explore whether weight gain was associated with a decrease in risk and weight loss was associated with an increase in risk. A total of 56,223 premenopausal participants in the Nurses' Health Study and 109,385 premenopausal participants in the Nurses' Health Study II were prospectively followed for up to 32 years and 18 years, respectively, and weight change since age 18 years was assessed biennially. The incidence of invasive breast cancer was assessed throughout follow-up. Weight loss of 5 kg or more since age 18, maintained for at least 4 years, was related to lower incidence of premenopausal breast cancer, compared to maintaining a stable weight, but this relation was of borderline statistical significance (covariate-adjusted HR = 0.75; 95% CI 0.52–1.09). Weight gain since age 18 years was also inversely related to breast cancer incidence among premenopausal women (covariate-adjusted p for trend = 0.01), but the association weakened after controlling for weight at age 18 and did not reach statistical significance (p for trend = 0.08). Although obesity and breast cancer among premenopausal women are inversely related, weight loss since age 18 years did not increase and weight gain did not significantly decrease the risk of premenopausal breast cancer among participants in the large prospective cohorts of NHS and NHS II.
In western countries, high body fatness is inversely related to breast cancer incidence among premenopausal women.1 The mechanisms underlying this perplexing association are not well understood. The question arises, whether weight loss after adolescence may increase the risk of premenopausal breast cancer and whether weight gain after adolescence decreases the risk of premenopausal breast cancer.
Conversely, a high body mass in adulthood increases breast cancer incidence among postmenopausal women.2 Weight gain since early adulthood and since menopause has been associated with an increase in postmenopausal breast cancer, while weight loss after menopause may reduce the risk.3 Weight loss lowers circulating estrogen levels among obese postmenopausal women, which may be the mechanism underlying the reduced breast cancer risk.4 Prior to menopause, women with a high body mass index (BMI) have lower total estradiol levels and higher free testosterone levels than their thinner counterparts.5–7 However, data on the effect of weight change on breast cancer incidence among premenopausal women are sparse.
In the Nurses' Health Study II (NHS II), we have previously found a more than 40% lower incidence of premenopausal breast cancer among women with a BMI at age 18 of 27.5 or larger compared to women with a BMI at age 18 between 20 and 22.4.1 We now investigated whether weight change since adolescence affects breast cancer incidence among premenopausal participants in the Nurses' Health Study (NHS) and NHS II.
Study population: Nurses' Health Study and Nurses' Health Study II
NHS was established in 1976, when 121,700 female registered nurses 30 to 55 years of age and living in 11 U.S. states completed a mailed questionnaire on their health status and on demographic, anthropometric and lifestyle factors. In 1989, 116,609 female registered nurses ages 25 to 42 years living in one of 14 U.S. states responded to a self-administered questionnaire about their medical histories and lifestyles, establishing the NHS II. Participants in both cohorts have since been followed with biennial questionnaires updating this information. Response rates since inception of the cohorts among living participants were 95% in NHS in 2008 and 94% in NHS II in 2007.
On every biennial questionnaire, women were asked whether their menstrual periods had ceased permanently and if so in which year their natural periods ceased and for what reason their periods ceased (natural, surgery, radiation/chemotherapy).
For this analysis, women were excluded at baseline (in 1976 for NHS and in 1989 for NHS II) if they were ineligible or died within the first 4 years of follow-up (n = 744), postmenopausal (natural, surgical or radiation/chemotherapy induced menopause) (n = 41,641), had a previous diagnosis of cancer other than nonmelanoma skin cancer (n = 6,041), were missing the date of diagnosis of invasive breast cancer (n = 688), or were missing information on weight at age 18 (n = 23,320) or height (n = 267), leaving 56,223 women from NHS and 109,385 women from NHS II for analyses.
Our study was approved by the Institutional Review Boards of the Brigham and Women's Hospital, Boston, MA, and the Harvard School of Public Health, Boston, MA.
Assessment of weight change
Information on height was obtained on the NHS baseline questionnaire in 1976, and weight at age 18 was queried in 1980. In NHS II, height and weight at age 18 were asked at baseline in 1989. The validity of recalled weight at age 18 and self-reported current height has previously been examined among 118 participants in the NHS II using records from physical examinations conducted at college or nursing school entrance.8 The correlation between recalled and measured past weight was 0.87. Mean BMI values were 21.6 kg/m2 for BMI calculated using recalled weight and were 22.1 kg/m2 using weight from medical records; the correlation was 0.84. Participants slightly under-reported weight at age 18 (mean difference = 1.4 kg).
Self-reports of weight were also validated in the Nurses' Health Study.9 Self-reported weight data were compared to standardized measurements taken ∼6 months apart by technicians who visited participants at their homes. Self-reported and measured weights were highly correlated at 97%.
Weight change since age 18 years was calculated as the difference between current weight (weight reported on the most recent questionnaire at each point in time) and weight at age 18 years. Women were categorized as having maintained weight since age 18 (lost or gained <2 kg), lost or gained weight since age 18. Weight change was considered stable if the direction of the weight change since age 18 remained the same for at least two consecutive questionnaire cycles.3 Weight reported in each weight cycle was evaluated relative to weight at age 18, not relative to the weight reported during the previous questionnaire cycle, hence a woman who continuously reports a weight 4 kg lower than her weight at age 18 throughout follow-up will always be counted as having a 4 kg weight loss. If the direction of the weight change remained stable for at least two consecutive questionnaire cycles but the amount changed relative to weight at age 18, a woman might contribute person-time to different weight change categories. A woman who gained weight relative to her weight at age 18 in one cycle and lost weight relative to her weight at age 18 during the next cycle did not contribute person-time until she maintained the loss relative to her weight at age 18 in the following cycle; 468,869 person-years were excluded due to unstable weight change and 600 incident breast cancers diagnosed during excluded person-years were censored. Within the stable weight change groups, women were categorized further by the amount of weight change since age 18: lost 2–4.9 kg, lost 5 kg or more, gained 2–4.9, 5–9.9, 10–14.9, 15–19.9, 20–24.9 or 25 kg or more. Women were also censored for follow-up intervals during which they reported a pregnancy.
Ascertainment of breast cancer cases
New cases of breast cancer were identified through biennial questionnaires mailed between 1976 and 2008 for NHS and 1989 and 2007 for NHS II. Deaths were reported by family members or by the Postal Service in response to the follow-up questionnaires, and the National Death Index was searched to investigate possible deaths among nonresponders. When a case of breast cancer was reported, we asked the participant (or next of kin for those who had died) for confirmation of the diagnosis and for permission to obtain relevant hospital records and pathology reports. Medical records were obtained for 97% of NHS cases and 82% of NHS II cases included in this analysis. Pathology reports confirmed breast cancer in more than 99% of women whose reports were reviewed, and for this reason, we included reported cases for which we were not able to obtain medical records. We restricted our study endpoint to invasive breast cancer. Cases of carcinoma in situ were censored at the time of diagnosis.
Assessment of covariate information
Information on potential confounding variables was assessed at baseline and during follow-up. Participants were asked for their date of birth, age at menarche and family history of breast cancer (in mother or sister) at baseline. History of benign breast disease, parity, age at first birth, alcohol consumption, oral contraceptive use and physical activity were assessed on baseline and subsequent questionnaires. History of breast cancer in the participant's mother and sisters was asked in 1976, 1982 and every 4 years since 1988 in NHS; in NHS II this information was obtained in 1989 and 1997. Data from subsequent questionnaires were used to update information on covariates by replacement with the new value for each individual in each time period.
Women were followed prospectively from the time they first reported their weight and height at baseline (1976 in NHS and 1989 in NHS II) until the end of follow-up (2008 in NHS and 2007 in NHS II). Women were included if they were premenopausal at baseline and censored from the analyses once they reached menopause; thus, only premenopausal cases were included in these analyses. Person-years of follow-up were calculated as the time from completion of the baseline questionnaire to the date of return of the final questionnaire (2008 in NHS and 2007 in NHS II), the date of diagnosis of invasive or in situ breast cancer, any other cancer (except non-melanoma skin cancer), death, loss to follow-up or reaching menopause, whichever occurred first. Women who had undergone a hysterectomy but had at least one ovary and were 47 years of age or younger if non-smoker or 45 years of age or younger if current smoker were considered premenopausal. Person-time of follow-up was excluded from the analysis if women were perimenopausal or if menopausal status was dubious or missing (172,750 person-years for NHS and 154,594 person-years for NHS II). Person-time was also excluded from the analysis if current weight was not reported on two or more consecutive questionnaires.
Data from NHS and NHSII were analyzed separately. A Cox proportional hazards model,10 stratified jointly by age in months and calendar year of follow-up at the beginning of each two-year questionnaire cycle, was used to calculate the hazard of developing invasive breast cancer associated with a particular level of weight change. All analyses were lagged by 4 years, therefore relating weight change (which had to have been stable for 4 years) to breast cancer newly diagnosed at least 4 years after stable weight change occurred to allow for a minimum latency period. Women diagnosed with breast cancer prior to 1984 in NHS and prior to 1997 in NHS II were excluded from the analysis because they had no opportunity to establish stable weight change prior to their cancer. Incident breast cancer cases diagnosed within 4 years of stable weight change were censored. Regression models were simultaneously adjusted for family history of breast cancer in first degree relative(s) (dichotomous), history of benign breast disease (dichotomous), height (continuous), age at menarche (≤10, 11, 12, 13, 14, 15+), parity (0,1, 2, 3, 4+), age at first birth (≤24 years, 25–30 years, >30 years), oral contraceptive use (never user, past user <5 years, past user ≥5 years, current user <5 years, current user 5–9 years, current user 10+ years), alcohol intake (none, <7.5 g/day, 7.5–15 g/day, 15–29 g/day, 30+ g/day) and physical activity (<3 Metabolic Equivalent Task (METS)/week, 3–9 METS/week, 9–17 METS/week, 18–26 METS/week, 27–41 METS/week, 42+ METS/week). We conducted separate analyses additionally adjusting for BMI at age 18. Covariate values were updated in the analysis whenever new information was obtained from the biennial questionnaire. The proportionality of relative hazards over time was examined by visual inspection.
Separate analyses were carried out for women with a low BMI (≤21) or high BMI (>21) at age 18 (the median BMI at age 18 value in NHS II) as well as estrogen receptor-positive (ER+) and progesterone receptor-positive (PR+) breast cancer and for estrogen receptor–negative (ER-) and progesterone receptor–negative (PR-) breast cancer.11
Because of the difference in calendar and follow-up time in the two cohorts, analyses were performed separately for each cohort, and the results were later combined using a fixed effects model weighting the two relative risk estimates by the inverse of the variance. Tests for heterogeneity were performed using the beta estimates for the ordinal weight change variable and employed to evaluate whether associations differed between the two cohorts; if heterogeneity had been identified, results would have been presented separately for the respective associations; however, no heterogeneity was found.
Trend tests were performed using the midpoint of the intervals. All tests of statistical significance were two-sided and a significance-level of 0.05 was used.
During 1,014,175 person-years of follow-up, 1,811 incident cases of invasive breast cancer were diagnosed among this premenopausal population of NHS and NHS II. The mean age at diagnosis for women with breast cancer was 49.1 years for NHS participants (5th percentile: 41.5 years; 95th percentile: 56.5 years) and 47.0 years for NHS II participants (5th percentile: 39.5 years; 95th percentile: 53.8 years). The simple mean difference between weight at age 18 and weight at the end of follow-up reflected a gain of 11.8 kg. Participant characteristics averaged over follow-up are presented in Tables 1 and 2. Compared to women who maintained a stable weight (lost or gained <2.0 kg) since age 18 years, those who gained weight were slightly older, had a higher BMI at age 18, were less likely to have a history of benign breast disease, were less likely to use oral contraceptives, were less physically active, and drank less alcohol (Tables 1 and 2). Women who lost weight since age 18 years were heavier at that age, were less likely to have a history of benign breast disease and were more likely to be nulliparous (Tables 1 and 2). Women who lost 5 or more kg had a mean BMI at age 18 of 25.6 kg/m2 in NHS and of 26.4 kg/m2 in NHS II; women who gained 5 or more kg had a mean BMI at age 18 of 21.0 kg/m2 in NHS and 21.0 kg/m2 in NHS II.
Compared to women who maintained a stable weight since age 18, women who lost 5 or more kg had a lower incidence of breast cancer with an age-adjusted hazard ratio (HR) of 0.61 (95% CI 0.44–0.85) (Table 3). After adjusting for family history of breast cancer, history of benign breast disease, height, age at menarche, parity, age at first birth, oral contraceptive use, alcohol intake and physical activity, the HR was 0.59 (95% CI 0.42–0.82), and after additional adjustment for weight at age 18 the HR was 0.75 (95% CI 0.52–1.09). No association was observed for weight loss between 2 and 4.9 kg and no trend emerged for weight loss. Weight gain was inversely associated with premenopausal breast cancer incidence (covariate-adjusted p for trend = 0.01), but this association weakened when accounting for absolute weight at age 18 (p for trend = 0.08 (Table 3). Among women who gained 25 kg or more since age 18, the HR for premenopausal breast cancer was 0.78 (95% CI 0.55–1.11).
Because weight change may have different effects depending on the body mass at age 18, we stratified our analysis by BMI at that age. Women whose BMI at age 18 years was above 21 kg/m2 and who lost 5 kg or more had an incidence of breast cancer not significantly different from those who maintained a stable weight (HR = 0.69; 95% CI 0.40–1.18); too few women started with a BMI of 21 kg/m2 or below and lost 5 kg or more to calculate a meaningful hazard ratio (Table 4). Associations between weight gain and breast cancer incidence did not differ significantly by BMI at age 18 years (p for heterogeneity = 0.1). While no association between weight gain and premenopausal breast cancer was observed for women with a BMI at age 18 of 21 kg/m2 or below, those who had a BMI at age 18 of above 21 kg/m2 and gained weight thereafter had a significantly lower incidence of breast cancer prior to menopause (p for trend = 0.02) [Table 4]. Women who gained 25 kg or more had a HR = 0.68 (95% CI 0.48–0.96).
Results did not differ among women with and without a family history of breast cancer (data not shown).
Separate analyses of weight change were also carried out for women whose breast cancer was estrogen receptor- and progesterone-receptor positive (ER+/PR+) and estrogen receptor- and progesterone-receptor negative (ER-/PR-). Weight loss since age 18 years was inversely related to both ER+/PR+ (HR for 5+ kg lost = 0.73; 95% CI 0.43−1.24) and ER−/PR− (HR for 5+ kg lost = 0.55; 95% CI 0.21−1.45) breast cancer, although statistical power was limited for ER−/PR− cases. No important association was observed between weight gain since age 18 and the incidence of either ER+/PR+ (p for trend = 0.4) or ER−/PR− breast cancer (p for trend = 0.8). No significant heterogeneity by receptor status was observed (p = 0.6). No important association was observed between weight gain since age 18 and the incidence of either ER+/PR+ or ER-/PR- breast cancer.
In this prospective cohort of women, we found weight loss and weight gain since age 18 associated with a marginally reduced incidence of breast cancer among premenopausal women, but these relations were not statistically significant.
The relation between trajectories of weight and breast cancer is complex and depends on the age at which weight and cancer are assessed. Obesity during childhood, adolescence and early adulthood is inversely associated with the risk of breast cancer among premenopausal women.2, 12–15 After menopause, a high BMI increases the risk of breast cancer.2 Sustained weight loss since age 18 and after menopause appears to have beneficial consequences for postmenopausal breast cancer risk,3 while weight gain increases incidence of breast cancer after menopause.3, 16, 17
The effect of weight change on premenopausal breast cancer risk has been considered in some previous studies. A previous analysis from NHS including 16 years of follow-up and 1,000 cases of premenopausal breast cancer suggested no association with weight loss since age 18, but few women lost weight.17 In the EPIC cohort, weight loss of more than 2 kg was associated with an increased incidence of breast cancer of borderline statistical significance, but statistical power was limited.18 In a case-control study in the United States, annual weight loss was inversely related to the risk of breast cancer; all information on weight was recalled.19 No association between weight loss and premenopausal breast cancer risk was found in a prospective cohort assembled in Norway and Sweden,15 in a large case-control study in Alberta, Canada,20 a nested case-control study in Hawaii21 and a case-control study in Washington State,22 while a positive relation with weight loss emerged in a German case-control study.23 In all studies, weight loss was assessed only once, or weight loss during the most recent follow-up interval was sufficient to qualify. Our study is unique in updating assessments of weight in 2-year intervals, in defining weight change as stable change that had to be maintained for at least 4 years, and in lagging the association between stable weight change and breast cancer diagnosis by 4 years.
In most studies on weight gain and premenopausal breast cancer, weight gain was not significantly associated with risk.18, 19, 21, 24 In other studies, an inverse association was apparent; however, in some of these studies, no adjustment was made for anthropometry at age 18 or 20 or at enrollment.20, 22 In the Scandinavian cohort study, the inverse link vanished after adjustment for weight at baseline.15 This is consistent with our finding that accounting for baseline BMI at age 18 is important when studying the association between weight change and premenopausal breast cancer. In the previous analysis of 16 years of NHS follow-up, we observed an inverse association of borderline statistical significance after adjustment for BMI at age 18.17
The mechanisms underlying the inverse association between BMI and premenopausal breast cancer remain to be elucidated. Key and Pike24 have suggested that anovulation among heavier women may play a role, resulting in decreased estradiol and progesterone levels. In a previous study, we conducted in NHS II, however, anovulation did not appear to be a primary explanation for the reduced risk among heavier women, since adjustment for menstrual cycle patterns, infertility due to ovulatory disorder, probable PCOS, or use of oral contraceptives did not even slightly attenuate the association with BMI.1
We modeled risk dependent on weight change stable for at least 4 years; this differs from a model which includes weight at age 18 and current weight while no information is provided on weight fluctuation between the two time points. Women who lost weight were heavier in adolescence than women who gained weight. It is possible, that the marginal inverse association observed with weight loss can be ascribed to the protection conferred by the overweight at age 18. Furthermore, weight loss may be a surrogate for an unmeasured lifestyle variable such as athletic activity. Individuals engaged in competitive athletics, such as high school team sports gain weight in muscle and bone density, but the extra weight is lost after returning to an average level of activity. Conversely, women heavier in adolescence may have adopted a more active lifestyle to intentionally loose weight. Women in this population who lost weight reported the highest levels of physical activity.
After adjusting for weight at age 18, the association with weight loss was attenuated. In previous studies, weight loss of 7 kg or more was association with a significant rise in Sex hormone binding globulin, a reduction in fasting glucose and fasting insulin, luteinizing hormone, free testosterone, and urinary cortisol excretion,25 and a significant decrease in both free and total estradiol as well as free and total testosterone.26, 27 Thus, reduced levels of free and total estradiol may mediate a reduction in breast cancer risk associated with weight loss. Conversely, high insulin plasma levels have been related to an increased risk of breast cancer in some studies;28, 29 thus, decreased insulin levels associated with weight loss may contribute to a reduction in breast cancer risk.
In our analyses, we allowed for a latency period of 4 years because it is unlikely that breast cancer cases newly diagnosed immediately after weight change that has been stable for 4 years are largely attributable to this change in weight. More likely, some time has to elapse during which the cancer grows. While the critical time window is unknown, the relation between body weight and breast cancer incidence changes throughout the lifecourse. While a high birthweight increases risk, a higher BMI in childhood until menopause is inversely correlated but a high BMI after menopause increases risk. These changes in associations suggest that weight may affect breast cancer risk within a few years.
Among potential confounding variables, physical activity deserves particular consideration. While we adjusted for physical activity in our analysis, women may not have been accurate in their estimation of regular exercise, which may have introduced residual confounding. If women who lost weight underreported physical activity, the observed decrease in breast cancer incidence may have been due to the exercise; if they overreported physical activity, the true effect of weight reduction on lowering breast cancer incidence may have been stronger. Conversely, if women who gained weight underreported physical activity, the true effect of weight gain on reducing breast cancer incidence may have been stronger; if they overreported physical activity, the observed decrease in breast cancer incidence may be at least partially explained by exercise.
In conclusion, despite an inverse relation between obesity and breast cancer among premenopausal women, weight loss since age 18 did not increase and weight gain since age 18 did not significantly decrease incidence of premenopausal breast cancer among participants in the large prospective cohorts of NHS and NHS II.
- 10Regression models and life-tables. J R Statist Soc 1972; 34: 187– 220..