Lifetime adult weight gain, central adiposity, and the risk of pre- and postmenopausal breast cancer in the Western New York exposures and breast cancer study
Version of Record online: 2 OCT 2006
Copyright © 2006 Wiley-Liss, Inc.
International Journal of Cancer
Volume 119, Issue 12, pages 2931–2937, 15 December 2006
How to Cite
Han, D., Nie, J., Bonner, M. R., McCann, S. E., Muti, P., Trevisan, M., Ramirez-Marrero, F. A., Vito, D. and Freudenheim, J. L. (2006), Lifetime adult weight gain, central adiposity, and the risk of pre- and postmenopausal breast cancer in the Western New York exposures and breast cancer study. Int. J. Cancer, 119: 2931–2937. doi: 10.1002/ijc.22236
- Issue online: 26 OCT 2006
- Version of Record online: 2 OCT 2006
- Manuscript Accepted: 26 JUN 2006
- Manuscript Received: 5 SEP 2005
- U.S. Army Medical Research Grants. Grant Numbers: DAMD17-03-1-1-0475, DAMD17-00-1-0417
- breast cancer;
- weight gain;
- case–control study
While there are quite consistent data regarding associations of body weight and postmenopausal breast cancer, there are now accumulating data that would indicate that weight gain in adult life is more predictive of risk than absolute body weight. There is, however, little known about the relative impact of timing of weight gain in adult life as well as other characteristics of the weight and breast cancer association that might provide insight into the mechanism of the observation. We conducted a population-based case control study of breast cancer (1996–2001), the Western New York Exposures and Breast Cancer Study. Included were 1,166 women with primary, histologically confirmed, incident breast cancer and 2,105 controls frequency-matched on age, race and county of residence. Unconditional logistic regression was used to estimate odds ratios and 95% confidence intervals. We found increased risk of breast cancer associated with lifetime adult weight gain among post- but not premenopausal women, and there was a 4% increase in risk for each 5 kg increase in adult weight. Further there was a tendency toward a stronger association for those with higher waist circumference and those with positive estrogen or progesterone status, and who had never used HRT. We also found an association with risk for weight gain since first pregnancy and for weight gain between the time of the first pregnancy and menopause, independent of body mass index and lifetime adult weight gain. Our results suggest that there are time periods of weight gain that have greater impact on risk, and that central body fat, receptor status and hormone replacement therapy may all affect the observed association. © 2006 Wiley-Liss, Inc.
Numerous epidemiologic studies of the relationship between body size and breast cancer risk have been conducted to examine its potential role as a modifiable risk factor, independent of dietary intake and physical activity.1, 2, 3, 4 Although our ability to explain the mechanism of the observed association is still limited, there is quite consistent evidence showing an association of indicators of body size and postmenopausal breast cancer. In particular, body mass index (BMI) and central adiposity have been shown to be associated with increased risk of post- but not premenopausal breast cancer.5, 6, 7, 8, 9
There are now accumulating data that would indicate that weight gain in adult life is more predictive of risk than absolute body weight or BMI.10, 11, 12, 13, 14, 15 There is, however, little known about the timing of weight gain in adult life as well as other characteristics of the weight and breast cancer association that might provide insight into the mechanism of the observation. There is interest in weight gain during particular periods of life, especially weight gain during periods of hormonal changes, such as pregnancy and menopause.4, 16, 17 Understanding of timing of weight gain in relation to risk could provide insight into the mechanism of the observed associations. In addition, there is evidence that central adiposity is associated with breast cancer risk7, 8, 9, 18, 19; interactions between weight gain and central adiposity are of interest, because of difference in metabolic activity depending on the location of fat disposition.
We examined here associations of lifetime adult weight gain with pre- and postmenopausal breast cancer, examining in particular the role of the timing of weight gain on postmenopausal breast cancer risk. We examined weight change between each decade as well as weight change at specific time points in a woman's life: around the time of women's first pregnancy and at menopause, as important time periods of hormonal change. Interrelationships between central adiposity and weight gain were also assessed to explain effects of adult weight gain on subsequent risk of postmenopausal breast cancer. Further, we evaluated whether effects of weight gain on postmenopausal breast cancer differed by estrogen and progesterone receptor status of tumors and by a women's use of hormone replacement therapy (HRT).
We conducted a population-based case control study of breast cancer (1996–2001), the Western New York Exposures and Breast Cancer Study (WEB Study). Eligible for the study as cases were all women diagnosed with primary, histologically confirmed, incident breast cancer, age 35–79 and resident of the 2 counties of the study area. All cases were interviewed within one year of diagnosis; most were interviewed within 3–6 months following diagnosis. Controls were randomly selected from the New York State Department of Motor Vehicles driver's license list (≤ aged 65) and the Health Care Finance Administration rolls (>65 years), frequency-matched to cases on age, race and county of residence. A total of 1,638 cases and 3,396 controls met our inclusion criteria of age 35–79, current residence in Erie or Niagara County in New York State, no previous cancer diagnosis other than nonmelanoma skin cancer. Response rates were 71% and 62% for cases and controls, respectively, among cases and controls for whom we could determine eligibility. All participants provided informed consent, and the protocol was approved by the Institutional Review Boards of the University at Buffalo and of all the participating hospitals.
Extensive in-person interviews and self-administered questionnaires were used to ascertain information on potential confounding factors and anthropometric measures, including lifetime weight. Participants were asked to recall their body weight for each decade of their lives from age 20 to 1 year before diagnosis for cases and 1 year before interview for controls. They were also asked to recall the amount of their weight gain during their first pregnancy, and to identify their body shape at the time of menarche from among 9 pictograms. Current height, weight and several measures of central adiposity (abdominal height, waist circumference and hip circumference) were measured by trained interviewers according to a standardized protocol. Waist circumference was measured by placing the tape around the smallest point between the top of iliac crest and the bottom of rib cage; hip circumference was measured by placing the tape around the hips at the biggest circumference point between the iliac crest and the crotch, and abdominal height was measured using a caliper on the participant in a recumbent position. All measures were to the nearest 0.1 cm. Three measurements were initially made for accuracy, and these were repeated until the 3 readings were all within 0.5 cm of each other. For the analyses reported here, current weight as reported weight 1 year before interview, and BMI based on measured height during interview and reported weight 1 year before interview were used. Weight 1 year before interview and measured weight were highly correlated (r = 0.91).
Estrogen receptor (ER) and progesterone receptor (PR) status information was abstracted from pathology reports. Receptor status was determined by either biochemical or immunoperoxidase assay; ER-positive tumors were those with at least an ER concentration values greater than 5 fmol/mg or with greater than 5% of the cells with ERs. We obtained ER and PR status for 646 cases of the 841 eligible postmenopausal breast cancer cases; for the remaining 23% (195) status was unknown either because women did not give us permission to examine their medical records (17) or because the receptor status of the tumor was not measured (178).
Lifetime adult weight change was calculated as the difference between reported weight at age 20 and weight 1 year before interview. Weight change was examined for a number of intervals that are potentially important: from the time of a first pregnancy and from the time of menopause to 1 year before interview, between the time of first pregnancy and menopause and during each decade of life from age 20.
All analyses were conducted stratified on menopausal status at diagnosis. Women were considered postmenopausal if their menses had ceased permanently and naturally. Among other women, participants were also considered postmenopausal if any of the following conditions were true: they had had a bilateral oophorectomy, they had had a hysterectomy without removal of the ovaries and they were older than 50, their menses had ceased permanently due to radiation or other medical treatment and they were older than 55.
Unconditional logistic regression was used to estimate odds ratios (OR) and 95% confidence intervals (CI). The cut points for the categorical analyses were derived from the distribution of controls; quartiles of weight gain were determined as well as another category of women who had not gained or who had lost weight during the time period of study. All models were adjusted for age at interview, education, age at first birth, age at menarche, previous benign breast disease, family history of breast cancer in a first degree relative and age at menopause and use of HRT (only for postmenopausal women). p for trend was determined by the p-value for the coefficient of the continuous exposure variable, while adjusting for covariates and excluding the group of women who had lost weight. Since weight change and BMI were highly correlated, we adjusted for BMI by including residuals of the regression of weight gain on BMI.20
While details of the study population have been published previously,21, 22 descriptive characteristics of cases and controls by menopausal status, including several anthropometric measures, were presented in Table I. Among premenopausal participants, these anthropometric measures were not different between cases and controls, with the exception of BMI at age 20 (p = 0.03) and body type at menarche (p = 0.04). Also weight gain since age 20 was not different between premenopausal cases and controls. Further we examined premenopausal breast cancer risk associated with BMI and weight change between age 20 and 1 year ago. As has been found by others, among the premenopausal women, BMI and lifetime weight gain were not associated with risk; adjusted OR was 0.75 (95% CI 0.49–1.16, p for trend = 0.18) for women in the highest quartile of BMI (>30.9 kg/m2) compared to the lowest (<22.2 kg/m2), while adjusted OR was 0.84 (95% CI 0.51–1.38, p for trend = 0.12) for women in the highest (>25 kg) compared to women who gained between 0 and 6.8 kg. Because the association of risk with weight gain was not seen among premenopausal women, further analyses were restricted only to postmenopausal participants.
|Premenopausal women (n = 935)||Postmenopausal women (n = 2,336)|
|Cases (n = 325)||Controls (n = 610)||p-value||Cases (n = 841)||Controls (n = 1,495)||p-value|
|Age (years)||44.9 ± 4.6||44.1 ± 4.6||0.01||63.0 ± 8.5||63.4 ± 8.9||0.30|
|Education (years)||14.0 ± 2.3||14.2 ± 2.2||0.08||13.3 ± 2.6||13.0 ± 2.3||<0.0001|
|Age at menarche (years)||12.5 ± 1.6||12.6 ± 1.6||0.76||12.6 ± 1.6||12.8 ± 1.7||0.02|
|Age at first pregnancy1 (years)||25.0 ± 5.1||25.8 ± 4.8||0.03||23.8 ± 4.7||23.5 ± 4.3||0.16|
|Age at menopause (years)||–||–||–||48.3 ± 5.4||47.4 ± 6.3||0.001|
|Time since first pregnancy1 (years)||19.9 ± 7.2||18.4 ± 6.8||0.01||39.5 ± 9.2||40.1 ± 9.3||0.19|
|Time between first pregnancy and menopause1 (years)||–||–||–||24.4 ± 6.8||24.1 ± 7.0||0.31|
|Height (cm)||164.0 ± 6.4||163.8 ± 6.3||0.63||161.8 ± 6.6||160.8 ± 6.1||<0.0001|
|BMI one year ago (kg/m2)||26.7 ± 6.6||27.2 ± 6.8||0.28||28.9 ± 6.1||28.4 ± 6.4||0.07|
|Weight at interview (kg)||73.1 ± 19.2||73.9 ± 19.0||0.52||75.7 ± 16.1||73.8 ± 16.3||0.01|
|Weight one year ago (kg)||71.7 ± 18.4||72.9 ± 19.2||0.35||75.6 ± 16.4||73.4 ± 17.0||0.003|
|Abdominal height (cm)||19.6 ± 3.3||19.6 ± 3.6||0.94||21.3 ± 3.2||20.9 ± 3.3||0.01|
|Waist circumference (cm)||84.4 ± 15.0||84.2 ± 15.2||0.86||91.3 ± 14.2||88.7 ± 14.0||<0.0001|
|BMI at age 20 (kg/m2)||20.9 ± 3.4||21.5 ± 4.0||0.03||21.0 ± 3.0||21.4 ± 3.2||0.01|
|Weight at age 20 (kg)||56.3 ± 10.2||57.6 ± 11.2||0.09||55.0 ± 8.5||55.2 ± 8.6||0.59|
|Weight gained during first pregnancy1||15.9 ± 8.6||14.9 ± 7.8||0.16||15.5 ± 10.8||15.1 ± 10.3||0.35|
|Weight gain since age 20 (kg)||15.4 ± 14.9||15.4 ± 15.8||0.35||20.6 ± 14.6||18.2 ± 15.4||<0.0001|
|Body type at menarche2||2.2 ± 1.2||2.4 ± 1.4||0.04||2.1 ± 1.2||2.1 ± 1.2||0.55|
|Relative with breast cancer (yes)||21%||10%||<0.0001||20%||14%||<0.0001|
|Use of hormone replacement therapy1 (yes)||–||–||–||54%||50.0%||0.09|
Selected anthropometric measures of postmenopausal participants are also shown in Table I; in general, anthropometric measures in early life, including weight at age 20, body type at menarche and weight gained during first full term pregnancy, were not different between cases and controls, while measures in later adult life, height, weight 1 year ago, measured weight, abdominal height and waist circumference were statistically different. Also, correlations between weight, BMI and weight change between time periods were examined; BMI 1 year ago was correlated with weight change between age 20 and 1 year ago (r = 0.85 and 0.86 for cases and controls, respectively).
Similarly, we found positive associations of postmenopausal breast cancer risk with most anthropometric measurements made at the time of interview when we examined association of various anthropometric measures in different time periods. Recent indicators of body size, including body weight 1 year before interview, height and BMI 1 year before interview, were all associated with increased risk of postmenopausal breast cancer; women with relatively higher body weight, height and BMI had an increased risk of 1.61 (95% CI 1.19–2.20, p = 0.02), 1.58 (95% CI 1.20–2.09, p = 0.001), 1.57 (95%CI 1.18–2.10, p = 0.02), respectively, when comparing highest to lowest quartile. We examined waist circumference as a measure of central adiposity; it was associated with postmenopausal breast cancer (adjusted OR 1.76, 95% CI 1.33–2.32). We also examined abdominal height in relation to risk; this measure of central adiposity was highly correlated with waist circumference (r = 0.85); the association with risk was similar to that for waist circumference (data not shown). Absolute body weight at first pregnancy and menopause were also associated with increased risk of postmenopausal breast cancer; women in the highest quartile of weight had a risk of 1.61 (95% CI 1.08–2.40) and 1.98 (95% CI 1.36–2.88), respectively, compared with the women in the lowest. Neither BMI at first pregnancy nor BMI at menopause was associated with increased risk of postmenopausal breast cancer. In addition, neither weight nor BMI at age 20 was associated with increased risk.
Risk associated with weight change between age 20 and 1 year ago for postmenopausal breast cancer are shown in Table II. An increased risk of breast cancer was found for postmenopausal women who gained more than 27.3 kg compared with women who gained between 0 and 9.1 kg (adjusted OR 1.71, 95% CI 1.23–2.37; p for trend = 0.05), even after controlling for BMI. Each 5 kg increase in the weight between age 20 and 1 year ago was associated with a 4% change in risk; OR 1.04 (95%CI 1.03–1.05) per 5 kg of weight gain in continuous form between 2 time points. Weight changes between decades and between several time periods with biological relevance to breast cancer were also examined. We found a positive association of weight change between several decades with postmenopausal breast cancer. There was an increase in risk of postmenopausal breast cancer associated with weight gain in the age periods 30–39 (OR 1.72, 95% CI 1.14–2.59), 40–49 (OR 1.92, 95% CI 1.17–3.15) and 60–69 (OR 1.94, 95% CI 1.21–3.10), but not in other decades of life; however, CIs for all decades overlapped. With adjustment for total lifetime weight change, ORs were attenuated and all CIs for weight gain during decades of life included the null.
|Cases||Controls||Adjusted OR1 (95% CI)||Adjusted OR2 (95% CI)|
|Weight change: age 20 to 1 year ago (kg) (n = 841 for cases; n = 1,495 for controls)|
|≤0||47||131||0.84 (0.53–1.33)||0.90 (0.56–1.45)|
|9.1–17.7||208||343||1.49 (1.10–2.01)||1.45 (1.06–1.96)|
|17.7–27.3||227||360||1.61 (1.19–2.17)||1.53 (1.12–2.08)|
|>27.3||222||331||1.86 (1.37–2.52)||1.71 (1.23–2.37)|
|Cases||Controls||Adjusted OR2 (95% CI)||Adjusted OR3 (95% CI)|
|Weight change: age 20–29 (kg) (n = 841 for cases; n = 1,495 for controls)|
|≤0||232||414||1.12 (0.82–1.53)||1.25 (0.92–1.71)|
|2.3–4.5||168||323||1.12 (0.82–1.53)||1.06 (0.78–1.45)|
|4.5–6.8||98||165||1.21 (0.83–1.77)||1.10 (0.75–1.61)|
|>6.8||166||263||1.40 (0.94–2.09)||1.08 (0.71–1.67)|
|Weight change: age 30–39 (kg) (n = 841 for cases; n = 1,495 for controls)|
|≤0||225||456||0.89 (0.66–1.19)||0.95 (0.71–1.27)|
|2.3–4.5||181||313||1.19 (0.87–1.61)||1.11 (0.82–1.52)|
|4.5–6.8||113||151||1.52 (1.06–2.18)||1.35 (0.93–1.96)|
|>6.8||143||229||1.72 (1.14–2.59)||1.32 (0.83–2.09)|
|Weight change: age 40–49 (kg) (n = 840 for cases; n = 1,494 for controls)|
|≤0||214||468||0.86 (0.56–1.31)||0.87 (0.57–1.32)|
|2.3–4.5||155||327||0.99 (0.65–1.53)||0.96 (0.63–1.48)|
|4.5–7.8||248||331||1.71 (1.13–2.59)||1.58 (1.02–2.43)|
|>7.8||163||256||1.92 (1.17–3.15)||1.60 (.91–2.79)|
|Weight change: age 50–59 (kg) (n = 790 for cases; n = 1,357 for controls)|
|≤0||218||441||0.99 (0.73–1.34)||1.05 (0.77–1.43)|
|2.3–4.5||182||253||1.52 (1.11–2.09)||1.39 (1.01–1.93)|
|4.5–9.1||70||145||0.88 (0.58–1.33)||0.76 (0.50–1.17)|
|>9.1||167||240||1.32 (0.87–1.99)||0.91 (0.56–1.48)|
|Weight change: age 60–69 (kg) (n = 497 for cases; n = 925 for controls)|
|≤0||170||331||0.99 (0.96–1.02)||1.09 (0.75–1.58)|
|2.3–4.5||82||131||1.33 (0.88–2.03)||1.19 (0.78–1.83)|
|4.5–8.3||44||101||1.06 (0.64–1.76)||0.90 (0.53–1.52)|
|>8.3||109||148||1.94 (1.21–3.10)||1.27 (0.72–2.27)|
|Weight change: age first pregnancy to 1 year ago (kg) (n = 693 for cases; n = 1,340 for controls)|
|≤0||43||150||0.75 (0.49–1.16)||0.74 (0.48–1.16)|
|7.7–14.1||175||303||1.48 (1.10–2.01)||1.50 (1.09–2.05)|
|14.1–22.7||174||303||1.53 (1.12–2.09)||1.57 (1.09–2.25)|
|>22.7||189||295||1.70 (1.22–2.37)||1.78 (1.08–2.94)|
|Weight change: age menopause to 1 year ago (kg) (n = 841 for cases; n = 1,495 for controls)|
|≤ 0||203||436||0.87 (0.64–1.19)||0.91 (0.66–1.25)|
|3.6–7.3||163||272||1.03 (0.74–1.43)||0.99 (0.71–1.38)|
|7.3–13.6||133||243||1.22 (0.86–1.73)||1.11 (0.76–1.61)|
|>13.6||201||290||1.58 (1.09–2.30)||1.27 (0.78–2.05)|
|Weight change: age first pregnancy to age menopause (kg) (n = 693 for cases; n = 1,340 for controls)|
|≤0||96||229||1.15 (0.80–1.66)||1.22 (0.84–1.76)|
|4.5–8.2||166||309||1.58 (1.14–2.20)||1.48 (1.06–2.06)|
|8.2–13.6||190||297||1.90 (1.36–2.66)||1.63 (1.13–2.33)|
|>13.6||150||262||1.91 (1.26–2.88)||1.34 (0.81–2.22)|
|Weight change: age 20 to age first pregnancy (kg) (n = 693 for cases; n = 1,340 for controls)|
|≤0||244||492||0.99 (0.73–1.35)||1.03 (0.76–1.41)|
|2.3–4.5||117||258||0.98 (0.70–1.36)||0.94 (0.67–1.32)|
|4.5–7.6||76||142||1.15 (0.78–1.72)||1.08 (0.72–1.61)|
|>7.6||139||212||1.52 (0.98–2.36)||1.25 (0.79–1.98)|
We also examined weight change around the time of women's first pregnancy and the time of menopause, and found increased breast cancer risk for weight change for weight gain between menopause and the present and for weight gain between first pregnancy and menopause. When weight change between first pregnancy and menopause was examined, we found positive associations; adult weight gain between these 2 hormonal time periods was strongly associated with the increased risk of postmenopausal breast cancer (OR 1.91, 95% CI 1.26–2.88, comparing highest to lowest quartile). We found similar associations of risk of postmenopausal breast cancer for weight change from first pregnancy to 1 year ago (OR 1.70, 95% CI 1.22–2.37), and from menopause to 1 year ago (OR 1.58, 95% CI 1.09–2.30). Again, CIs for these time periods overlapped and these results overlapped with those for the decade analysis. We further adjusted these ORs for weight gain between age 20 and 1 year ago; ORs were diminished except for the one for weight change since first pregnancy that remained of similar magnitude. On the other hand, weight change between age 20 and first pregnancy was not associated with the risk. We did not observe an association of reported weight gain during the first pregnancy with breast cancer risk (data not shown).
Additionally, breast cancer risk associated with weight change between age 20 and 1 year ago were analyzed by the time since menopause. We found that adult weight gain was more strongly associated with increased risk of breast cancer among postmenopausal women with a longer time since menopause; there was an increase in risk associated with weight gain for those women in the strata more than 20 years, and 11–20 years since menopause (OR 2.35, 95% CI 1.24–4.46, p-trend = 0.05 and OR 2.00, 95% CI 1.10–3.64, p-trend = 0.01, respectively), but not in less than 10 years (OR 1.25, 95% CI 0.74–2.12, p for trend = 0.59).
We also examined the relationships between weight change and central adiposity with breast cancer risk. ORs for risk associated with weight change between age 20 and 1 year ago, stratified by the categories defined by the median of waist circumference were calculated (Table III). Adult weight gain and central body fat were correlated (r = 0.68); there was more weight gain for both cases and controls among those with more central body fat. Adult weight gain was associated with increased risk of breast cancer only among women with waist circumference above the median. Although a test for interaction was not statistically significant (p = 0.42), there was a nonmonotonic increase in the group of women with waist circumference above the median, but there was no association with waist circumference below the median.
|Weight change (kg)||Below median (<88 cm)||Above median (≥88 cm)|
|Cases (n = 366)||Controls (n = 745)||Adjusted OR1 (95% CI)||Cases (n = 475)||Controls (n = 750)||Adjusted OR1 (95% CI)|
|≤0||38||105||0.86 (0.50–1.48)||9||26||1.61 (0.49–5.20)|
|9.1–17.7||126||215||1.22 (0.84–1.76)||82||128||2.48 (1.16–5.28)|
|17.7–27.3||67||131||1.10 (0.68–1.76)||160||229||2.39 (1.15–4.95)|
|>27.3||12||17||1.31 (0.52–3.27)||210||314||2.33 (1.11–4.90)|
Finally, we evaluated whether the association of weight gain with risk was different by ER and PR status. Adult weight gain was strongly associated with increased risk of postmenopausal breast cancer among ER- or PR-positive tumors, and statistically significant trends were observed (p = 0.001, Table IV). Weight change was not associated with breast cancer risk among ER- or PR-negative tumors. While there was some tendency for the risk of postmenopausal breast cancer in relation to adult weight gain to be more strongly associated with risk among never- than among ever-users of HRT, CIs overlapped for the upper category in both strata.
|Weight change (kg)||ER+||ER−|
|Cases (n = 510)||Controls (n = 1,495)||Adjusted OR1 (95% CI)||Cases (n = 136)||Controls (n = 1,495)||Adjusted OR1 (95% CI)|
|≤0||30||131||1.03 (0.58–1.82)||10||131||1.31 (0.51–3.33)|
|9.1–17.7||124||343||1.60 (1.09–2.35)||30||343||1.48 (0.78–2.80)|
|17.7–27.3||138||360||1.86 (1.27–2.73)||42||360||1.62 (0.86–3.05)|
|>27.3||149||331||2.42 (1.62–3.61)||30||331||1.19 (0.58–2.43)|
|Weight change (kg)||PR+||PR−|
|Cases (n = 389)||Controls (n = 1,495)||Adjusted OR1 (95% CI)||Cases (n = 257)||Controls (n = 1,495)||Adjusted OR1 (95% CI)|
|≤0||20||131||1.05 (0.52–2.12)||20||131||1.13 (0.58–2.18)|
|9.1–17.7||94||343||2.07 (1.32–3.27)||60||343||1.14 (0.72–1.82)|
|17.7–27.3||107||360||2.39 (1.52–3.76)||73||360||1.33 (0.84–2.11)|
|>27.3||123||331||3.14 (1.96–5.04)||56||331||1.21 (0.73–2.01)|
|Weight change (kg)||Never-users of HRT||Ever-users of HRT|
|Cases (n = 385)||Controls (n = 716)||Adjusted OR2 (95% CI)||Cases (n = 447)||Controls (n = 717)||Adjusted OR2 (95% CI)|
|≤ 0||20||64||0.75 (0.34–1.63)||27||62||1.03 (0.56–1.89)|
|9.1–17.7||96||160||1.69 (1.04–2.75)||109||171||1.32 (0.88–1.97)|
|17.7–27.3||97||164||1.86 (1.15–3.03)||127||182||1.38 (0.91–2.08)|
|>27.3||123||182||2.00 (1.22–3.27)||96||128||1.52 (0.96–2.41)|
We investigated the association of lifetime adult weight gain and pre- and postmenopausal breast cancer, in particular weight changes at different time points in a woman's life. As has been reported previously,15, 23, 24 we observed increased risk of breast cancer associated with BMI and lifetime adult weight gain among post- but not premenopausal women; there was a 4% increase in risk for each 5 kg increase in adult weight. Further, we found that lifetime adult weight gain was associated with risk of postmenopausal breast tumors that were ER- or PR-positive and there was a tendency toward a stronger association for those with higher waist circumference and who had never used HRT. Current weight and waist circumference were associated with risk. We did not find any association with risk for measures of body size at the time of menarche, weight at age 20 or weight gain during the first pregnancy. In relation to timing of weight gain, we did find an association of weight gain since first pregnancy, and weight gain between the time of the first pregnancy and menopause, independent of BMI 1 year ago and lifetime adult weight gain. Weight gain since menopause was also associated with risk. Adult weight gain was more associated with risk for women whose menopause was further in the past than for those who had experienced it more recently.
Timing of weight gain has not generally been addressed adequately in previous studies of weight change and breast cancer, and such timing may provide a clue on etiologic role of adult weigh gain in relation to breast cancer risk. We found positive associations with risk for weight gain during the 30s and 40s, and between the time periods of hormonal changes. However, the mechanism for these findings is not known. It could be that weight gain is an indicator of hormonal environment and that the factors that lead to increased body weight also increase risk. It may also be that energy balance at these time points is particularly significant. We also found positive associations of adult weight gain with risk for those with positive ER or PR status and for never users of HRT, consistent with the findings of others.13, 23, 24, 25, 26, 27 These findings also lend credence to the notion that the association of weight gain with risk of postmenopausal breast cancer is related to steroid hormone metabolism. It may also be that insulin resistance and impaired glucose metabolism related to increased BMI explain our findings, at least in part. Insulin is a powerful mitogenic agent. In cell culture, insulin induces dose-dependent growth response in breast cancer cell lines acting via insulin receptor.28, 29 Insulin may play a role in tumor promotion by upregulation of ovarian steroid secretion.30 Intraportal insulin levels influence IGF-1 bioavailability,31 and recent studies suggest an etiological role of IGF-I in the development of breast cancer.32, 33, 34 There is also epidemiologic evidence of a close association between major alteration in glucose metabolism and breast cancer risk; in a prospective study there was a doubling of breast cancer risk for women who had a diagnosis of diabetes at baseline.35
Our findings provide some suggestion that only among women with greater central adipose is weight gain associated with risk; it has been suggested that central adipose tissue is more metabolically active than peripheral adipose.18 In addition, there has been speculation that timing of weight gain may have different effects on the location of fat disposition.36, 37 Further, greater central adiposity may reflect differences in steroid hormones including testosterone.38, 39 It could be that changes in glucose metabolism associated with increased central body fat explain the observed associations, providing additional evidence linking central adiposity to altered glucose metabolism. Previous studies have generally shown increased risk of postmenopausal breast cancer in relation to various measures of central adiposity in most,7, 8, 9, 19 but not in all studies.26, 27 Waist–hip ratio was the commonly used measure of central adiposity, but some recent studies have used waist circumference that has been shown to be a stronger predictor of breast caner risk than waist to hip ratio.9, 18 These interactions between central adiposity and weight gain during the period of hormonal changes may also provide evidence of the importance of metabolic changes in relation to breast cancer risk.
To further evaluate timing of weight gain on the risk, we examined effects of early life body-size indicators on breast cancer risk. Recent studies have shown some evidence that obesity in childhood is protective,14, 40 and higher growth rates during adolescence may be an independent risk factor of breast cancer in adulthood.41, 42 When we asked about a crude measure of body shape at the time of menarche, we did not find any association of this measure with risk. As have others, we found a protective effect of higher BMI at age 20; women in the highest quartile had a risk of 0.73 (95% CI 0.55–0.97), supporting the hypothesis that higher weight in early adulthood reduces the risk of breast cancer in adulthood. In addition, height, possibly a marker of exposure to nutritional factors in early life, was associated with increased postmenopausal breast cancer risk.
In our study, there were a small number of women who reported losing weight during their adult lives or during one of the time periods. For most measures, there were nonsignificant increases in risk in this group. We also looked at risk associated with weight gain when this group was included in the referent. Findings were similar to those presented where the absence of weight gain or loss was the referent. Further, because of the possibility of weight loss in late stage breast cancer patients around the time of diagnosis, we examined results restricted to early stage breast cancer; findings were not different from those for all breast cancer cases shown here (data not shown). Additionally, all models in our study were adjusted for the frequency matching factor of age. Controls were also frequency matched on race and county of residence. Models including those as control variables did not materially affect the point estimates.
Strengths of this population-based case–control study include the assessment of adult weight change at numerous time periods. We were able to assess various indicators of weight change and central adiposity, especially during periods of hormonal change in a woman's life, and to assess effect modification of this association by stratifying analyses by hormone receptor status and the use of HRT. Because of the comprehensive interview data regarding lifetime body weight, we were able to examine lifetime patterns in relation to risk.
There are several limitations that need to be considered in interpretation of these results. These include those common to the case–control study design: recall and selection bias as well as misclassification in the self-reported measures. To assess selection bias, we obtained general information in a brief phone interview from nonparticipants who were willing to give information, and found that nonparticipants were not different from those participants for several characteristics, including diet. However, we did not query nonparticipants regarding anthropometry. There may be recall bias in the lifetime self-reported weight. However, we used the same methods for both cases and controls, and there was no particular emphasis on weight-related questions. Interviewers were blinded as to case and control status. In addition, reported weight 1 year before interview had a high correlation with measured weight at the time of interview, and was similar for cases and controls (0.92 and 0.91 for cases and controls, respectively). For the lifetime weight change, we used reported weight for age 20 and for 1 year before interview. We were concerned that for cases the current measured weight might be affected by their disease or by treatment. In fact, there were few differences. For waist circumference we relied on the interviewer measurements. Another issue is that of multiple comparisons. Because of the large number of results for this study, it may be that some are by chance alone. There has been considerable discussion of the issue of multiple comparisons in the epidemiologic literature.43 Clearly these findings need to be replicated in other studies before causal inferences can be drawn.
This study supports epidemiologic evidence showing an increased risk of breast cancer associated with adult weight gain among postmenopausal women. We did not find that any 1 decade of life was more strongly associated with risk. We did find some suggestion that weight gain during the 30s and 40s, weight gain since a woman's first pregnancy and weight gain since menopause, especially for women with a longer time since menopause may all be of importance in relation to postmenopausal breast cancer risk. We found no association of weight gain during pregnancy and risk. Our results are consistent with evidence from previous studies, demonstrating interactions of weight gain and central adiposity measures, and suggesting that weight gain during periods of hormonal change are associated with higher risk. Further, the findings of an association with risk limited to ER- or PR-positive tumors also appear to provide evidence that the mechanism of an association of body weight with risk includes a hormonal etiology. Timing of adult weight gain appears to be of importance in relation to risk; evaluation of weight gain during times of hormonal changes and in relation to central adiposity should be evaluated in other research settings.
- 3International Agency for Research on Cancer (IARC). Handbooks of cancer prevention: weight control and physical activity. Lyon: IARC Press, 2002.
- 20Nutritional epidemiology. Oxford: Oxford University Press, 1998..