- © 2005 by American Society of Clinical Oncology
Weight, Weight Gain, and Survival After Breast Cancer Diagnosis
- From the Channing Laboratory, Department of Medicine, Brigham & Women's Hospital, and Harvard Medical School; and Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, MA
- Address reprint requests to Candyce Kroenke, ScD, MPH, Channing Laboratory, 181 Longwood Ave, 3rd floor, Boston, MA 02115; e-mail: candyce.kroenke{at}channing.harvard.edu
Abstract
Purpose To determine whether weight prior to diagnosis and weight gain after diagnosis are predictive of breast cancer survival.
Methods Patients included 5,204 Nurses' Health Study participants diagnosed with incident, invasive, nonmetastatic breast cancer between 1976 and 2000; 860 total deaths, 533 breast cancer deaths, and 681 recurrences (defined as secondary lung, brain, bone, or liver cancer, and death from breast cancer) accrued to 2002. We computed the change in body mass index (BMI) from before to the first BMI reported ≥ 12 months after the date of diagnosis. Cox proportional hazards models were used to evaluate associations of categories of BMI before diagnosis and of BMI change with time to event. We stratified by smoking, menopausal status, and breast cancer–related variables.
Results In multivariate-adjusted analyses, weight before diagnosis was positively associated with breast cancer recurrence and death, but this was apparent only in never smokers. Similarly, among never-smoking women, those who gained between 0.5 and 2.0 kg/m2 (median gain, 6.0 lb; relative risk [RR], 1.35; 95% CI, 0.93 to 1.95) or more than 2.0 kg/m2 (median gain, 17.0 lb; RR, 1.64; 95% CI, 1.07 to 2.51) after diagnosis had an elevated risk of breast cancer death during follow-up (median, 9 years), compared with women who maintained their weight (test for linear trend, P = .03). Associations with weight were stronger in premenopausal than in postmenopausal women. Similar findings were noted for breast cancer recurrence and all-cause mortality.
Conclusion Weight and weight gain were related to higher rates of breast cancer recurrence and mortality, but associations were most apparent in never-smoking women.
INTRODUCTION
Overweight at the time of breast cancer diagnosis has been linked frequently to poorer survival in most studies,1-9 although some investigators have found a weak or no relationship between weight and recurrence or mortality.10-14
Overweight and obese postmenopausal women have been observed to have higher estrogen and androgen concentrations compared with lighter weight women.1,15 In obese women, there may be enhanced conversion in the adipose tissue of androgens to estrogens.1,16,17 In addition, obesity has been related to later stage at diagnosis and larger tumor size—both important predictors of subsequent survival.9,18,19 Other mechanisms through which obesity may influence survival include decreased levels of sex hormone-binding globulin, increased insulin and insulin-like growth factors,20 lower treatment efficacy,21 and the correlation with black race and lower socioeconomic status.22-25
Many women report gaining weight as a result of their breast cancer diagnosis,26 due in part to the effects of certain treatment regimens, most notably chemotherapy.27,28 Many women are also concerned about weight gain from tamoxifen, although several placebo-controlled trials have not supported an association between tamoxifen and weight gain.29-31 Possible mechanisms for weight gain include fatigue and reduced physical activity,32,33 reductions in lean body mass and resting energy expenditure,26 and increased ingestion of food as a means to cope or because of treatment-related increases in appetite.33
Given a possible adverse effect of overweight before diagnosis on breast cancer survival, weight gain associated with breast cancer diagnosis may also have an adverse effect on survival. Few studies have explored weight gain after breast cancer diagnosis and survival, and these findings are mixed,7,34-37 due in part to limited power with small sample sizes.34-36 In the largest study of weight gain and survival to date (n = 545), Camoriano et al7 found that women who were relapse-free by 60 weeks after study enrollment and who gained more than the median weight had a higher risk of relapse and death than those who gained less than the median weight. Additional research exploring how survival is influenced across the range of weight gain is needed.
We hypothesized that Nurses' Health Study (NHS) participants with breast cancer who gained weight after diagnosis would have a shorter survival (both disease-free and overall survival) than women who maintained their weight. We also evaluated associations of weight before breast cancer diagnosis with survival to provide a context for our findings.
METHODS
NHS
The NHS is a prospective study of 121,700 US female nurses, 30 to 55 years of age at baseline in 1976. At baseline and during biennial follow-up periods, participants provided detailed lifestyle and medical history information through a mailed questionnaire.
In this study, we included women who were diagnosed with invasive breast cancer between 1976 and 2000 and who answered questions on height in 1976 and pre- and postdiagnosis weight. We excluded women with in situ disease or metastatic breast cancer at diagnosis. Metastatic cancer was defined as having a physician-confirmed diagnosis of metastatic breast cancer. Women with four or more positive nodes but lacking a complete metastatic work-up were also excluded because of concerns for occult metastatic disease. A complete metastatic work-up consisted of a negative chest x-ray (or chest computed tomography), bone scan, and liver function tests (or liver scan), or documentation from a treating physician that the patient did not have metastatic disease. We also eliminated all recurrences and deaths as a result of breast cancer that occurred within a year of diagnosis to eliminate patients who likely had metastatic disease but were not classified as such by our initial definition. A total of 5,204 women were included in the study.
Data Collection
Measurement of breast cancer.
In the NHS, incident breast cancer was ascertained by biennial mailing of the questionnaire to participants. For any report of breast cancer, written permission was obtained from study participants to review their medical records. Physicians, blinded to exposure information from questionnaires, subsequently reviewed medical records and pathology reports. Overall, 99% of self-reported invasive breast cancers for which medical records are obtained have been confirmed.
Measurement of breast cancer recurrence.
If, after having a breast cancer diagnosis, women reported a second cancer on a routine NHS follow-up, we assumed that breast cancer had recurred if she reported lung, liver, bone, or brain cancer, as these are the most common sites of recurrence. We also defined recurrence to include women who did not report a second cancer but who died as a result of breast cancer. In this definition, we assumed disease recurred 2 or more years before death. If patients died within 2 years of diagnosis, the date of recurrence was defined as the date of death. Before exclusions, numbers of patients with recurrent breast cancer calculated in this manner were fairly similar to the numbers expected given the recurrence rates found in a large (N = 5,569) trial of radiation treatment in early-stage breast cancer.38 We did not include either ipsilateral or contralateral breast recurrences in our definition of recurrence. Because our definition of recurrence has not been validated, we conducted sensitivity analyses redefining recurrence to include only those women reporting a second cancer.
Measurement of mortality.
Ascertainment of deaths in the NHS cohort included reporting by the family or postal authorities. In addition, names of persistent nonresponders were searched in the National Death Index. Date of death was ascertained from death certificates, supplemented as needed with medical records. More than 98% of deaths in the NHS cohort have been identified by this method.39 Mortality in this study was defined as breast cancer mortality if cause of death was listed on the death certificate as breast cancer. To decrease the possibility of misclassification of cause of death, death certificates, medical records, and information from family members were reviewed by physicians. All-cause mortality was defined as death as a result of any cause.
Measurement of change in body mass index, pre- to postdiagnosis.
Prediagnosis body mass index (BMI) was computed from height in 1976 and weight reported on the biennial survey prior and most recent to the diagnosis of breast cancer. If information on weight was missing, we used the weight reported on the previous biennial survey. If weight was missing at both time points, women were excluded from study. Postdiagnosis BMI was computed from weight reported on the survey after breast cancer diagnosis if reported weight was provided ≥ 12 months after diagnosis, to allow for completion of treatment. If data were missing, we used information on weight from the next follow-up, excluding those with weight missing at both time points after diagnosis. Self-reported weight in this cohort has been validated with a correlation of 0.99 between self-reported and actual weight.40
Change in BMI was computed by subtracting prediagnosis from postdiagnosis BMI. We did not update postdiagnosis BMI in computation of BMI change to avoid problems with interpretation, given that weight loss after diagnosis may signify disease recurrence.
Statistical Analyses
In analyses of weight before diagnosis, women were categorized as having BMI less than 21, 21 to 22, 23 to 24, 25 to 29, and 30+ kg/m2. Women with BMI of 21 to 22 kg/m2 comprised the reference. In analyses of change in weight, women were categorized as losing more than 0.5 kg/m2, maintaining weight (within ± 0.5 kg/m2 of original BMI; reference), gaining between 0.5 and less than 2.0 kg/m2, and gaining 2.0 kg/m2 or more. Categories were created to ensure adequate distribution across categories and to explore both how moderate and substantial gains in weight would influence results.
Using linear regression, we regressed potential confounding variables against categories of BMI change, adjusted for continuous age (Table 1).
Selected Characteristics Across Categories of Weight Change Among 5,204 Women With Breast Cancer From the Nurses' Health Study
We employed Cox proportional hazards models (SAS PROC PHREG; SAS Institute, Cary, NC) for failure-time data to evaluate associations of categories of baseline BMI and change in BMI before and after breast cancer diagnosis with time to recurrence, breast cancer death, or all-cause mortality.41,42 We evaluated associations with all-cause mortality to overcome limitations with possible misclassification of cause of death when determining disease-specific mortality.43 Relative risk estimates were obtained by exponentiating the fitted β to a particular model, and 95% CIs were obtained by exponentiating the 95% confidence bounds of β. Person-months of follow-up were counted from the date of breast cancer diagnosis until the date of recurrence or date of death (depending on the analysis) or end of follow-up, whichever came first. Follow-up ranged from 2 to 26 years, with a median follow-up of 9 years. We evaluated the P value (test for linear trend for the top three categories of weight change only), setting those reporting weight loss to missing, given that we expected the overall relationship of weight change and mortality to be U-shaped rather than linear (ie, we expected women experiencing weight loss to have a higher mortality rate than women who maintained weight).
Results adjusted for age and prediagnosis BMI (the latter only in analyses of change in weight) were compared with those obtained, adjusting for these factors and multiple factors shown to be predictive of survival after breast cancer diagnosis in the NHS cohort including oral contraceptive use, menopausal status, age at menopause, use of hormone replacement therapy, protein intake, tumor size, nodal involvement, chemotherapy, and hormonal therapy. We adjusted for parity and age at birth using Rosner's and Colditz's birth index, which enables fine adjustment for parity and age at each birth with the inclusion of a single, continuous variable.44 We also adjusted for time elapsed between baseline BMI assessment and time of diagnosis, as well as time between diagnosis and postassessment.
Because of the metabolic effects of cigarette smoking on body weight and strong association between cigarette smoking and all-cause mortality, we stratified by smoking status—never versus past or current smokers. We evaluated models stratified by variables including tumor size (≤ 2 v > 2 cm), nodal status (node-negative v node-positive), stage using the standard American Joint Committee on Cancer staging criteria (stage 1, 2, or 3),45 estrogen receptor (ER) status (ER-negative, defined as both ER and progesterone receptor [PR] -negative, versus ER-positive, defined as either ER- or PR-positive), treatment regimen (chemotherapy, tamoxifen), menopausal status (premenopausal v postmenopausal), current postmenopausal hormone use (postmenopausal women only), and overweight status at baseline (defined as BMI ≥ 25 kg/m2). Interaction terms were computed for categoric BMI change and dichotomous versions of each of these variables and were evaluated with Wald tests. We used the regression method by Greenland and Finkle46 to include women in analyses with missing information on continuous variables.
Finally, we conducted several sensitivity analyses. We performed analyses with and without eliminating all cases of prior cancer. Although in main analyses, we excluded women who had a recurrence or died within 1 year of diagnosis, we also conducted analyses including these women to ensure comparability to previous studies. As previously indicated, we defined recurrence in two different ways. We also performed analyses with different cut points (1.5, 2.0, and 2.5 kg/m2), dividing the top two categories of change in BMI. All tests of statistical significance were two sided.
RESULTS
Of the 5,204 women with invasive breast cancer, 860 died (533 as a result of breast cancer) and 681 had a breast cancer recurrence. In the two categories of BMI gain (Table 1), median weight gain was 6 and 17 lb. Women who gained weight were slightly younger at the time of diagnosis compared with those who lost or maintained weight. Reproductive and cancer-related variables did not vary linearly by categories of BMI change, although women with the largest gains in weight tended to be more likely to have received treatment with chemotherapy or tamoxifen and have later stage tumors. Protein consumption was positively related to change in BMI (Table 1).
Among all breast cancer survivors, body weight before breast cancer diagnosis was weakly and positively related to breast cancer death and all-cause mortality but not recurrence (Table 2). However, in analyses stratified by smoking or menopausal status, positive associations of weight and breast cancer mortality were apparent and strong in never-smoking or premenopausal women but not in ever-smoking or postmenopausal women (test for interaction by smoking status, P = .06; test for interaction by menopausal status, P = .08; Table 3). Similarly, associations with breast cancer recurrence and total mortality were stronger in never smokers or premenopausal women compared with ever smokers or postmenopausal women (data not shown).
Relative Risk of All-Cause Mortality, Breast Cancer Death, and Recurrence by Category of Weight Prior to Diagnosis Among 5,204 Women With Breast Cancer From the Nurses' Health Study
Multivariate-Adjusted* RR of Breast Cancer Death by Category of Weight Prior to Diagnosis Among 5,204 Women With Breast Cancer From the Nurses' Health Study
In analyses stratified by both smoking and menopausal status, among never smokers, we found a positive association of baseline weight with recurrence and mortality in premenopausal women. In postmenopausal women, we found a weaker, significant, positive association of baseline weight and all-cause mortality (test for linear trend, P = .01), a suggestion of a positive association with breast cancer survival, and no association with breast cancer recurrence (data not shown). Correspondingly, we found a significant interaction of baseline weight and menopausal status among never smokers for recurrence (test for interaction, P = .01) but not for breast cancer or all-cause mortality (data not shown). Among smokers, a significant, positive association of weight before diagnosis and survival was present in premenopausal (test for linear trend, P = .03) but not postmenopausal women (data not shown). However, interactions among smokers were nonsignificant.
In multivariate-adjusted analyses, weight gain was associated with an increased risk of recurrence, breast cancer death, and total mortality, but only among never-smoking women (Table 4). Results were similar in age and baseline BMI-adjusted analyses. On adjusting for covariates, the association of weight gain with breast cancer survival was somewhat attenuated, particularly in the highest category of BMI gain, and this was due primarily to adjustment for nodal involvement. In an analysis of smokers, there was no apparent effect of weight gain and breast cancer recurrence or mortality (Table 4).
Relative Risk of All-Cause Mortality, Breast Cancer Death, and Recurrence, by Category of Weight Change Among 5,204 Women With Breast Cancer From the Nurses' Health Study
In other stratified analyses of weight gain, there were significant interactions by nodal status (test for interaction, P = .05) and tumor size (test for interaction, P = .04) with results stronger among women with node-negative cancer or smaller cancers. There was also a significant interaction between change in BMI and baseline overweight status for breast cancer death (Table 5). The association of weight gain and breast cancer mortality was evident only in women who were normal weight (BMI < 25 kg/m2) but not overweight (BMI ≥ 25 kg/m2) at baseline (test for interaction, P = .02).
Stratified Relative Risk of Breast Cancer Death by Categories of BMI Change Among 5,402 Women With Breast Cancer From the Nurses' Health Study
The number of ER-negative breast cancers was small, so we lacked power to fully examine this interaction. There were no other important differences in other stratified analyses (eg, by treatment factors or menopausal status; data not shown).
Results did not vary when we excluded women with a prior history of other cancers or when we altered the definition of recurrence. We found similar results with various cut points for BMI change (data not shown). Finally, we found similar patterns of association whether or not we included women who had a recurrence or died within the first year of diagnosis.
DISCUSSION
We found that high body weight before breast cancer diagnosis was associated with poorer survival after diagnosis. This relationship was stronger in never smokers and premenopausal women. In addition, women who gained weight after breast cancer diagnosis had an elevated risk of both recurrence and mortality, adjusting for weight before diagnosis. This risk increased with increasing weight gain. Again, this relationship was most apparent in never smokers and also among women with earlier stage disease or those who were normal weight before diagnosis. This study suggests that both weight before diagnosis and weight gain after diagnosis may be associated with poorer survival.
Camoriano et al7 found poorer survival among women gaining more versus less than the median weight. We found increasing risk with increasing weight gain compared with women who maintained their weight. The stronger positive associations of weight or weight gain and survival in premenopausal than postmenopausal women in our study are consistent with their results and with previous results showing no significant effect of weight on breast cancer prognosis in postmenopausal women12,13 and a positive association in premenopausal women.9,47 Also consistent with our data, prior research has often shown that the effect of overweight on prognosis is strongest among women with less severe disease.5,48-51
Most prior research has also shown that women who were overweight at the time of diagnosis have worse prognosis after breast cancer diagnosis.1 However, more recent studies suggest little or no association of weight at diagnosis with subsequent survival.4,10,11,21,52 These differences among studies could be related to sample size as well as the characteristics of the population under study, including the proportion of smokers or postmenopausal women.
Importantly, no prior studies have stratified by smoking status. Given the important influence of smoking on weight and mortality, the failure to adjust fully for smoking may obscure the influence of weight or weight gain on breast cancer recurrence and mortality. Smokers are leaner. Smoking may both promote the formation of less biologically active estrogens53,54 and attenuate postcancer gains in weight. Smoking may also decrease the risk of breast cancer (and possibly recurrence) among women with a genetic predisposition.55 Furthermore, it is well known that smokers are at increased risk of all-cause mortality compared with nonsmokers. Because of competing risk of death from alternate causes such as cardiovascular diseases, these women may be less likely to die as a result of breast cancer. However, because of the potential effect of smoking on the minimization of weight gain, associations may also be attenuated in analyses of all-cause mortality.56
In a study of body weight and mortality, Manson et al56 stratified by smoking status and found that the association of weight and mortality was direct among never smokers and J-shaped among past and current smokers and in the total sample (adjusted for smoking), suggesting confounding that could not be properly adjusted for in analyses without stratifying by this variable. Therefore, by mixing smokers and nonsmokers, previous studies may have underestimated the impact of weight and weight change.
Additional weight gain after a breast cancer diagnosis appeared to confer no additional risk to survival among women who were already overweight before diagnosis, however. Overweight women tend to be diagnosed with later stage cancer with more adverse tumor characteristics than normal-weight women, which may explain the stronger association between weight gain and breast cancer mortality in normal-weight women. Alternatively, there may be little additional adverse hormonal impact of weight gain on breast cancer survival in already overweight women.
Strengths of this study include a large sample size and the ability to stratify by breast cancer–specific factors. We were also able to prospectively explore weight change with survival and stratify additionally by smoking, which enabled clearer interpretation of findings.
Study limitations include reliance on self-reported weight. However, the correlation between self-reported and actual weight has been high (r = 0.99) in the NHS, and self-reported BMI has been an excellent predictor of many disease outcomes and mortality.56 We also lacked complete information on breast cancer treatment and recurrence, and the method of identification of recurrence has not been validated. However, a subanalysis excluding women without complete treatment information did not substantially alter results. Furthermore, by including breast cancer deaths in the definition of recurrence, there was a lower likelihood of results being influenced by detection bias. Results were also similar whether or not we included breast cancer deaths among recurrences and for breast cancer and all-cause mortality.
In summary, weight before diagnosis was related to worse prognosis, especially among never smokers and premenopausal women. Similarly, weight gain was related to higher rates of breast cancer recurrence and mortality in never-smoking women, and furthermore in women who were normal weight at diagnosis or had earlier stage cancers. The relationship between weight, weight gain, and breast cancer survival may be more complex than originally considered. Consistent with results from this study and guidelines provided by the American Cancer Society,57 women should attempt to maintain weight after breast cancer diagnosis.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
Acknowledgments
We thank Lisa Li for her invaluable programming assistance.
Footnotes
-
Supported by National Institutes of Health grants AG/CA14742 and CA87969.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
- Received January 14, 2004.
- Accepted September 9, 2004.
