breast neoplasms; cohort studies; hereditary diseases; obesity; risk

Transcription

1 American Journal of Epidemiology Copyright 2002 by the Johns Hopkins Bloomberg School of Public Health All rights reserved Vol. 155, No. 3 Printed in U.S.A. Waist/hip Ratio, Family History, and Breast Cancer Sellers et al. Interaction of Waist/Hip Ratio and Family History on the Risk of Hormone Receptor-defined Breast Cancer in a Prospective Study of Postmenopausal Women Thomas A. Sellers, 1 Jenny Davis, 1 James R. Cerhan, 1 Robert A. Vierkant, 1 Janet E. Olson, 1 V. Shane Pankratz, 1 John D. Potter, 2 and Aaron R. Folsom 3 The authors previously reported an interaction of waist/hip and family history on the risk of breast cancer in the Iowa Women s Health Study. Here they reexamine this association based on 9 additional years of followup, stratifying on tumor receptors for estrogen and progesterone. Data on risk factors and family history of breast cancer were ascertained in The occurrences of breast cancer and estrogen receptor/progesterone receptor were determined through the Iowa Surveillance, Epidemiology, and End Results registry. s were elevated with increasing weight and body mass index and decreasing body mass index at age 18 years, but they did not vary by family history. There was no association with height, waist circumference, or waist/hip. A linear trend of increasing risk with increasing waist/hip was observed among family history-positive women (p = 0.06) but not among family history-negative women (p = 0.87). This apparent interaction (p = 0.09) was examined by estrogen receptor or progesterone receptor status. When stratified on family history and estrogen receptor, no clear patterns were evident. In contrast, family history-positive women in the upper quintile of the waist/hip were at 2.2-fold greater risk of progesterone receptor-negative tumors compared with those in the lowest quintile ( : 0.9, 5.8). Thus, the previously reported interaction between family history and waist/hip is still (weakly) evident and appears to reflect risk for progesterone receptor-negative tumors. Am J Epidemiol 2002;155: breast neoplasms; cohort studies; hereditary diseases; obesity; risk There is considerable evidence that the risk of postmenopausal breast cancer is increased by obesity (1). There is increasing recognition, however, that the distribution of body fat may be an additional risk factor. Abdominal adiposity, which can be measured in epidemiologic studies as the of waist circumference to hip circumference (2), appears to be more metabolically active than fat deposited peripherally. In addition, abdominal adiposity is associated with increased levels of serum androgens, insulin, and C- peptide and decreased levels of sex hormone-binding globulin (3 5). There are consistent data from prospective cohort studies that abdominal adiposity increases the risk of postmenopausal breast cancer (6 9). There are some discrepancies, however, as to whether the waist circumference or waist/hip is the more relevant risk factor (9). In an earlier report from the Iowa Women s Health Study cohort, we Received for publication February 16, 2001, and accepted for publication July 25, Department of Health Sciences Research, Mayo Clinic and Mayo Clinic Cancer Center, Rochester, MN. 2 Cancer Prevention Program, Fred Hutchinson Cancer Research Center, Seattle, WA. 3 Division of Epidemiology, University of Minnesota, Minneapolis, MN. Reprint requests to Dr. Thomas A. Sellers, Mayo Clinic, 200 First Street SW, Rochester, MN ( sellers@mayo.edu). described an interaction between a high waist/hip and family history of breast cancer on the risk of disease (10). Specifically, the risk associated with a high waist/hip was primarily limited to those women in the cohort who had reported a family history of the disease. A subsequent report (11) suggested that the interaction was stronger among women with a family history of breast and ovarian cancer, a combination of malignancies more likely to be influenced by genetic predisposition. Waist circumference was not examined in either of those two previous Iowa Women s Health Study reports. Breast cancer is clearly a heterogeneous disease. It has been hypothesized that one way to identify different etiologic pathways is to examine the association of known and suspected risk factors within strata defined by the presence or absence of hormone receptors (12, 13). We previously reported data from the Iowa Women s Health Study on the association of measures of obesity with breast tumors stratified on the joint distribution of estrogen receptor and progesterone receptor status (12). The results suggested that a high body mass index and a high waist/hip were primarily associated with an increased risk of progesterone receptor-positive tumors. The purpose of the current report is to update our earlier publication (10), based on only 5 years of follow-up and 492 cases of breast cancer, of an interaction between the waist/hip and family history. Currently, 13 years of 225

2 226 Sellers et al. follow-up of the cohort have been completed, with 1,874 incident cases of breast cancer. The significantly increased number of events afforded the opportunity to explore the interaction within strata defined by tumor receptors for estrogen or progesterone. The association of obesity with postmenopausal breast cancer has been shown to be strongest with estrogen receptor-positive/progesterone receptor-positive tumors (12, 14). These reports, plus the suspicion that abdominal adiposity influences the risk of breast cancer through perturbations in hormone levels, led us to hypothesize that interactions of family history and waist/hip would be strongest for tumors that express the estrogen and progesterone receptor. MATERIALS AND METHODS Definition of cohort Detailed methods of the Iowa Women s Health Study have been published elsewhere (15). Briefly, in January 1986, we mailed a 16-page questionnaire to 98,029 eligible women aged years who resided in Iowa. We randomly selected potential participants from the state of Iowa s driver s license list, and the 41,836 respondents (42.7 percent response rate) form the cohort under study. The rates of breast cancer among responders and nonresponders were virtually identical (15). Risk factor assessment The questionnaire solicited information on factors known or suspected to be relevant to breast cancer risk, including pregnancy history, menstrual history, physical activity, smoking history, and alcohol use. Family history of breast cancer was assessed for mothers, sisters, daughters, maternal and paternal grandmothers, and maternal and paternal aunts. For the purpose of this report, data only on first-degree relatives reported at baseline were used. No data on family size or the age at onset of breast cancer in relatives were collected. Family history of breast cancer was reassessed on the third follow-up survey; the proportion of respondents who reported a family history was 15.7 percent, slightly higher than the baseline prevalence of 12.3 percent. Because participation on the third follow-up was not associated with family history at baseline (80.4 percent participation vs percent participation, respectively), this probably reflects a real increase. The body mass index (weight (kg)/height (m) 2 ) was calculated from self-reported current height and weight. To assess body fat distribution, a paper tape was enclosed along with instructions to have a friend measure the circumference of the waist (1 inch (2.54 cm) above the umbilicus) and hips (maximal protrusion). Measurements obtained by this method have been shown to be both accurate and reliable in this cohort (2). The waist/hip was calculated as the of these two circumferences. Exclusion criteria For the analyses presented here, we excluded women at baseline if they were not postmenopausal (n 569), had had a mastectomy or partial breast removal (n 1,870), or had any cancer other than skin cancer at baseline (n 2,293). These exclusions left a total of 37,105 women eligible for follow-up. Follow-up We mailed follow-up questionnaires in 1987, 1989, 1992, and 1997 to establish vital status and change of address. Through linkage with the Iowa death certificates, supplemented by linkage to the National Death Index, we identified nonrespondent members of the cohort who were deceased. Vital status is estimated to be unknown for less than 1 percent of the cohort. We ascertained cancer incidence through the State Health Registry of Iowa, a part of the National Cancer Institute s Surveillance, Epidemiology, and End Results Program (16). Field representatives routinely visited hospitals and clinics in and around Iowa. For cancer patients who were Iowa residents at the time of diagnosis, information including personal identifiers, tumor size and grade, estrogen receptor and progesterone receptor status, and extent of disease was recorded. Annually, we matched by computer a list of cohort members and the records of Iowans with incident cancer in the Health Registry using combinations of first, last, and maiden names; zip code; birth date; and Social Security number. Through December 31, 1998, corresponding to 13 years of follow-up, we identified 1,874 cases of breast cancer (1,643 invasive and 231 in situ) among the cohort at risk. Analytical approach The length of follow-up for each woman in the study was calculated as the time from completion of the baseline questionnaire until the date of breast cancer diagnosis, date of move from Iowa, or date of death. If none of these events applied, follow-up continued through December 31, s and 95 percent s were calculated using Cox proportional hazards regression models. We first assessed the overall association of breast cancer separately for each of the following anthropometric measures: height, weight, body mass index, waist/hip, body mass index at age 18, and waist circumference. Each variable was categorized by quintiles, and rate s were calculated using the lowest category as the referent group. We then calculated tests for trend by ordering the categories from lowest to highest and including this variable in a proportional hazards regression model as a linear variable. Next, we evaluated whether a family history of breast cancer (defined as having a mother, sister, or daughter diagnosed with breast cancer) modified the association between the anthropometric measures and breast cancer risk by checking the statistical significance of the interaction term between these measures and family history in the Cox models. For ease of interpretation, the original set of main effects and interactions was then multiplied by a contrast matrix that allowed for direct comparison of anthropometrics within each level of family history but that did not change the overall fit of the model. Thus, the risk of breast cancer in

3 Waist/hip Ratio, Family History, and Breast Cancer 227 all subsequent tables for a given anthropometric measure is presented with the lowest quintile as the referent category for each family history category. The main effects of anthropometrics were then evaluated separately for estrogen receptor-positive and -negative tumors and progesterone receptor-positive and -negative tumors, respectively. In these analyses, the outcome variable was incident receptor status-specific breast cancer, and all other types of breast cancer were considered censored observations. Next, we determined if risk s for the exposures of interest differed according to receptor status using a competing risk form of Cox proportional hazards analysis (17). This approach allowed us to model and test the interaction between a given risk factor (modeled as a covariate) and receptor status (included as a stratum variable). Separate models were fit for each anthropometric measure, first comparing estrogen receptor-positive tumors with estrogen receptor-negative tumors and then comparing progesterone receptor-positive tumors with progesterone receptor-negative tumors. For all Cox models, survival was modeled as a function of age, because age is a better predictor of breast cancer risk than is length of follow-up time in this study (18). Each model included the following potential confounding variables as covariates: education level, age at menarche, age at menopause, oral contraceptive use, hormone replacement therapy, parity, age at first birth, alcohol use, smoking status, and physical activity. The anthropometric variables included as potential confounders were selected after screening for colinearity using Spearman s rank correlations. Specific adjustments for each anthropometric exposure are given as footnotes to the tables. All statistical tests were two sided, and all analyses were carried out using the SAS (SAS Institute, Inc., Cary, North Carolina) and Splus (Mathsoft, Inc., Seattle, Washington) software systems. RESULTS Through 13 years and more than 415,000 person-years of follow-up, 1,874 breast cancers were identified in the cohort at risk. Women who reported at baseline a family history of breast cancer in a first-degree relative (12.3 percent) were at 1.47-fold greater risk (95 percent : 1.30, 1.66) than were women without such a family history. Table 1 shows the distribution of anthropometric risk factors by family history of breast cancer. All distributions were similar by family history status. The association of anthropometric risk factors with subsequent occurrence of breast cancer by family history is shown in table 2. All rate s were adjusted for age and other risk factors. Height and waist circumference were not associated with increased risk, regardless of family history. Risks were increased with greater weight and body mass index and decreased with greater body mass index at age 18 years, but the patterns and magnitude were similar across family history categories. Conversely, the association between waist/hip and postmenopausal breast cancer appeared to differ by family history (p 0.09). For the total cohort, the rate for the highest versus the lowest quintile of waist/hip was 1.10 (p 0.31). Among women without a family history of postmenopausal breast cancer, there was no evidence of increased risk; rather, the modest increase in risk observed among the entire cohort appeared to be limited to the subset of women who reported a family history of breast cancer at baseline. The rate for the highest quintile compared with the lowest was 1.55 (95 percent : 1.04, 2.32), with a marginal linear trend across all categories (p 0.06). Analyses were then performed to determine if the associations of anthropometric variables with breast cancer risk differed by the hormone receptor status of the tumor. Women for whom the receptor status was unknown were included as a separate category. The majority of cases with missing data on receptor status were because of the small size of the tumors. As shown in table 3, the risks associated with current weight, body mass index, and body mass index at age 18 years were primarily limited to tumors that were positive or missing for estrogen and progesterone receptors. Height, which was unrelated to risk in the total at-risk cohort, demonstrated a weak positive association with estrogen receptor-negative and progesterone receptor-negative tumors. Both the waist and waist/hip, which showed no overall association with risk in the entire cohort, were associated only with tumors lacking data on receptor status. The next set of analyses explored whether the observed interaction of waist/hip and family history of breast cancer was specific to a receptor-defined subset of tumors. Because of small numbers, models were constructed for estrogen receptor and progesterone receptor separately. Figure 1 shows the calculated rate s for breast cancer by waist/hip, family history, and estrogen receptor status. No clear patterns were evident. In contrast, when the data were stratified according to progesterone receptor status, a clearer pattern emerged (figure 2). For the progesterone receptor-positive tumors and the family history-negative/ progesterone receptor-negative tumors, there appeared to be no association between waist/hip and the risk of breast cancer. However, women with a family history of breast cancer who were also in the highest quintile of waist/hip were at 2.2-fold elevated risk of developing progesterone receptor-negative breast cancer compared with women with a family history of breast cancer and low waist/hip (95 percent : 0.87, 5.76). DISCUSSION The present study suggests that the interaction of waist/hip and family history of breast cancer that was previously reported on the Iowa Women s Health Study cohort (10) is still evident. However, 9 additional years of follow-up attenuated the magnitude of the interaction and the risk associated with a high waist/hip among women with a family history of breast cancer. We had hypothesized that analyses stratified on tumor hormone receptor status would indicate an association primarily with estrogen receptor-positive or progesterone receptor-positive breast cancer. Contrary to expectations, the results provide evidence suggesting that any excess risk associated with a high waist/hip among

4 228 Sellers et al. TABLE 1. Distribution of baseline anthropometric risk factors by family history of breast cancer, Iowa Women s Health Study, 1986 No family history Positive family history Anthropometric risk factor No. % No. % Height (inches)* >66 8,610 4,522 5,281 8,036 5, , , Weight (pounds) >174 Body mass index (kg/m 2 ) Waist/hip >0.90 Body mass index at age 18 years (kg/m 2 ) >22.90 Waist (inches) > ,344 6,229 6,746 5,817 6,383 6,245 6,473 6,267 6,307 6,227 5,579 5,712 7,181 5,983 6,941 6,090 6,448 6,268 6,295 6,286 6,198 6,268 6,654 6,198 6,102 women with a family history of breast cancer is a reflection of elevated risk of progesterone receptor-negative tumors. In the early years of cohort follow-up, data on hormone receptor status were slightly less likely to be included in the pathology reports than in later years. In particular, the proportion of cases with missing data on receptor status was 29 percent between 1986 and 1990, 21 percent between 1991 and 1994, and 20 percent between 1995 and Inspection of table 3 shows that the risks associated with a high waist/hip were most evident for tumors missing receptor status, which tended to be smaller or in situ tumors: 55 percent of the cases with missing data on tumor receptor , * Metric cutpoints are 157 cm, cm, cm, cm, and >168 cm. Metric cutpoints are 58.1 kg, kg, kg, kg, and >78.9 kg. Metric cutpoints are 76 cm, cm, cm, cm, and >99 cm status had tumors of less than 5 mm or of unknown size. Because pathologists and Surveillance, Epidemiology, and End Results abstractors were blinded to family history and waist/hip, it is difficult to imagine that this reflects any sort of selection bias. Although the literature on abdominal adiposity and breast cancer risk is fairly consistent, the observation of an interaction with abdominal adiposity and family history of breast cancer has not been reported outside the Iowa Women s Health Study cohort. There are no obvious explanations for this discrepancy, although one must certainly consider the possibility that the interaction observed among the Iowa Women s

5 Waist/hip Ratio, Family History, and Breast Cancer 229 TABLE 2. Multivariate-adjusted* rate of postmenopausal breast cancer by anthropometric risk factors and family history of breast cancer, Iowa Women s Health Study, 1986 Anthropometric factor Height (inches) >66 No. of cases Total no. of women 7,760 4,088 4,860 7,260 4,1 No family history , , , , 1.22 p value test for trend p = 0.55 No. of cases Total no. of women , Positive family history , , , , 1.69 p value test for trend p = 0.46 Weight (pounds) >174 Body mass index (kg/m 2 ) Waist/hip >0.90 Body mass index at age 18 years (kg/m 2 ) >22.90 Waist circumference (inches) > ,7 5,701 6,107 5,274 5,754 5,700 5,911 5,711 5,692 5,585 5,156 5,218 6,559 5,403 6,2 5,543 5,892 5,735 5,753 5,676 5,705 5,708 6,092 5,605 5, , , , , , 1.50, 1., , , , , , , , , , , , , , 1.48 p < p < p = 0.87 p < p = , , , 2.39, , , , , , , , , , , , , , , , , 1.75 p = 0.02 p = 0.05 p = 0.06 p < p = 0.75 * All analyses include adjustment for education, age at menarche, age at menopause, oral contraceptive use, hormone replacement therapy, parity, age at first birth, alcohol use, smoking status, physical activity level, and family history of breast cancer. Additional adjustment for weight, waist/hip, and body mass index at age 18 years. Metric cutpoints are 157 cm, cm, cm, cm, and >168 cm. Additional adjustment for height, waist/hip, and body mass index at age 18 years. Metric cutpoints are 58.1 kg, kg, kg, kg, and >78.9 kg. Additional adjustment for height, body mass index, and body mass index at age 18 years. Metric cutpoints are 76 cm, cm, cm, cm, and >99 cm. Additional adjustment for height, waist/hip, and body mass index. Health Study cohort is simply chance. That the interaction has persisted through 13 years of follow-up and appears to have greater specificity for a particular subtype of breast cancer (progesterone receptor-negative tumors) argues against chance as the explanation. Because of the limited amount of epidemiologic research on this topic, it might be helpful to examine the biologic plausibility of the current findings to help disentangle chance statistical associations from truth. A logical starting point in the search for biologic plausibility is a review of the role of progestins and the proges-

6 TABLE Multivariate-adjusted* rate of postmenopausal breast cancer by anthropometric risk factors and hormone receptor status, Iowa Women s Health Study, Anthropometric value Height (inches) >66 Current weight (pounds), >174 Body mass index (kg/m 2 ), Waist/hip # >0.90 Body mass index at age 18 years (kg/m 2 )** >22.90 Waist circumference (inches)# >39.00 Estrogen receptor positive (n = 1,043) , , 1.32, , 1.39, , , , , , , , , , , , , , , 0.49, , , , , 1.44 Estrogen receptor negative (n = 232) , , , , , , , , , , , , , , , , , , , , , , , , 2.09 Missing estrogen receptor (n = 445) , 0.62, 0.71, , , , , , , , , , , , , 2.24, , , , , , , , , 2.39 Progesterone receptor positive (n = 993) , , , , , , 2.41, , , 1.65, , , , , , 0.83, , , 0.61, , , , , 0.78, 1.57 Progesterone receptor negative (n = 362) , , , , , , , , , ,, , , , , , , , , , , , , , 1.72 Missing progesterone receptor (n = 519) , , , , , , , , , , , , , , , , , , 0.52, , , , , , 2.11 * All analyses include adjustment for education, age at menarche, age at menopause, oral contraceptive use, hormone replacement therapy, parity, age at first birth, alcohol use, smoking status, physical activity level, and family history of breast cancer. Additional adjustment for weight, waist/hip, and body mass index at age 18 years. Metric cutpoints are 157 cm, cm, cm, cm, and >168 cm. Additional adjustment for height, waist/hip, and body mass index at age 18 years. Metric cutpoints are 58.1 kg, kg, kg, kg, and >78.9 kg. Test for interaction between estrogen receptor positive and estrogen receptor negative: p = 0.05; test for interaction between progesterone receptor positive and progesterone receptor negative: p = Test for interaction between between progesterone receptor positive and progesterone receptor negative: p = # Additional adjustment for height, body mass index, and body mass index at age 18 years. Metric cutpoints are 76 cm, cm, cm, cm, and >99 cm. ** Additional adjustment for height, waist/hip, and body mass index. ownloaded from by guest on 02 October Sellers et al.

7 Waist/hip Ratio, Family History, and Breast Cancer 231 FIGURE 1. Adjusted rate of postmenopausal breast cancer according to family history of breast cancer, waist/hip, and estrogen receptor status, Iowa Women s Health Study, FHx, family history; ER, estrogen receptor; +, positive;, negative. terone receptor in breast cancer. Mitotic activity of normal breast epithelial cells is greatest during the luteal phase of the menstrual cycle. The observation that estrogen receptors are down regulated but progesterone receptors are not during the luteal phase (19) suggests that progestins may be especially important for mitogenicity. The progesterone receptor belongs to the superfamily of intracellular receptors that mediate the nuclear effects of steroid hormones, thyroid hormone, and vitamins A and D (20). It contains a highly conserved DNA-binding domain, a hormone-binding domain, and a variable N-terminal domain. The human progesterone receptor exists as two isoforms reflecting dis- FIGURE 2. Adjusted rate of postmenopausal breast cancer according to family history of breast cancer, waist/hip, and progesterone receptor status, Iowa Women s Health Study, FHx, family history; PR, progesterone receptor; +, positive;, negative.

8 232 Sellers et al. tinct estrogen-inducible promoters within the gene: hpr-b (114 kda) and hpr-a (94 kda) (21). Experimental evidence suggests that only hpr-a is capable of opposing transcriptional activity mediated by the estrogen receptor (22). Thus, absence of functional hpr-a could lead to diminished ability to modulate epithelial prolifen induced by estrogen. Under the assumption that family history represents a crude surrogate for differences in genes, one potential factor that may underlie the family history waist/hip interaction is genetic variation within the progesterone receptor itself. A 306-base pair insertion in intron 7 of the human progesterone receptor (PR) gene has been identified (23) that is in linkage disequilibrium with a mutation in exon 4 that causes an amino acid change in the hinge region of the receptor. There is some evidence that this polymorphism is associated with decreased breast cancer risk among women younger than age 50 years, but only among women without a family history of the disease (24). However, there were too few postmenopausal subjects in that study to help inform interpretation of the current results. Moreover, the literature regarding the polymorphism is inconsistent. Two studies have found no association with breast cancer risk and that, when loss of heterozygosity is detected in the tumor tissues of allele carriers, it is not the wild-type allele that is lost (25, 26). Finally, it is unknown whether the genetic polymorphism is associated with lack of expression of the receptor in tumor tissue. Two additional candidates that could account for the apparent gene environment interaction include BRCA1 and BRCA2. There is emerging evidence that women with inherited mutations in either of these genes are more likely to present with progesterone receptor-negative tumors than are women with sporadic breast cancer (27, 28). Within the current study, we performed ad hoc analyses of additional features of their family history that may be indicative of an inherited mutation in BRCA1 in the 62 cases with a family history of breast cancer and a high waist/hip. Compared with women who developed progesterone receptor-positive tumors (n 46), there was some evidence that those who developed progesterone receptor-negative tumors (n 16) were more likely to have a family history of ovarian cancer in a first-degree relative (12.5 percent vs. 4.4 percent), to have a family history of breast and ovarian cancer (18.8 percent vs. 6.7 percent), and to have a relative affected with breast cancer before the age of 45 years (38.5 percent vs percent). We recently reported evidence of linkage to the BRCA1 locus among a small subset of 13 breast cancer families selected from the Iowa Women s Health Study (29). Interestingly, the evidence for linkage was evident only among families in which the proband was in the upper median (n six families) of waist/hip (unpublished data). However, complete sequencing of the BRCA1 gene failed to reveal any truncating mutations. Given this inconsistency within our study population, the low frequency of BRCA1 mutations, and the limited contribution of known BRCA1 mutations to postmenopausal breast cancer, it seems unlikely that BRCA1 interacts with a high waist/hip to increase risk. Additional research is needed to verify this assertion. Another plausible biologic mechanism for the observed interaction may relate to differences in hormone metabolism (30). It is important to reiterate that a high waist/hip was not associated with elevated risk among women without a family history of breast cancer. One way to interpret this is that, whatever the consequence of a high waist/hip may be on hormone or growth factor levels, it would appear to trigger malignancy only among a subset of those with a high waist/hip. Among obese postmenopausal women, this may occur through genetic variation in the capacity to convert androgens to estrogens in adipose tissue. The gene encoding the enzyme for this activity, Cyp19, is polymorphic in the population. Conversely, the effects could be downstream from estrogen formation, through metabolism to catecholestrogens. Experimental evidence suggests that certain catecholestrogens can form depurinating DNA adducts and lead to initiation or progression of malignancy (30). Several candidate genes with known functional polymorphisms that impact on the ability to detoxify activated estrogens have been studied (such as COMT, GSTM1, GSTT1, and GSTP1). It would be important to explore interactions of these polymorphisms with abdominal adiposity in future studies. In an earlier report from this cohort on risk factors for breast cancer by hormone receptor status (12), we observed three patterns of risk: progesterone receptor positive; estrogen receptor positive, progesterone receptor negative; and estrogen receptor negative, progesterone receptor negative. It would have been interesting to use the same categories in the current report, but because of stratification by family history, small numbers of cases in some categories precluded estimation of rate s with precision. Continued follow-up of the cohort should allow these analyses to be done in the future. In conclusion, the current report finds continued evidence for an interaction between family history of breast cancer and a high waist/hip on the risk of postmenopausal breast cancer. The elevated risk appears to be particularly elevated for progesterone receptor-negative tumors, but a larger sample size will be needed to verify this. Given the lack of consistency in the literature regarding this interaction, it is critical that additional work should explore the possible biology and should study variation at particular loci rather than rely on family history. ACKNOWLEDGMENTS This study was supported by grant R01 CA39742 from the National Cancer Institute to A. R. F. Technical assistance was provided by Shannon Seeger. REFERENCES 1. van den Brandt PA, Spiegelman D, Yaun SS, et al. Pooled analysis of prospective cohort studies on height, weight, and breast cancer risk. Am J Epidemiol 2000;152: Kushi LH, Kaye SA, Folsom AR, et al. Accuracy and reliability of self-measurement of body girths. Am J Epidemiol 1988; 128: Seidell JC, Cigolini M, Deurenberg P, et al. Fat distribution,

9 Waist/hip Ratio, Family History, and Breast Cancer 233 androgens, and metabolism in nonobese women. Am J Clin Nutr 1989;50: Kaye SA, Folsom AR, Soler JT, et al. Associations of body mass and fat distribution with sex hormone concentns in post-menopausal women. Int J Epidemiol 1991;20: Olson JE, Anderson KE, Cerhan JR, et al. An investigation of the biological basis of an interaction of abdominal fat distributuion and family history of breast cancer. A nested study of sisters in the Iowa Women s Health Study (United States). Cancer Causes Control 2000;11: Folsom AR, Kaye SA, Prineas RJ, et al. Increased incidence of carcinoma of the breast associated with abdominal adiposity in postmenopausal women. Am J Epidemiol 1990;131: Ballard-Barbash R, Schatzkin A, Carter CL, et al. Body fat distribution and breast cancer in the Framingham Study. J Natl Cancer Inst 1990;82: Kaaks R, Van Noord PA, Den Tonkelaar I, et al. Breast-cancer incidence in relation to height, weight and body-fat distribution in the Dutch DOM cohort. Int J Cancer 1998;76: Huang Z, Willett WC, Colditz GA, et al. Waist circumference, waist:hip, and risk of breast cancer in the Nurses Health Study. Am J Epidemiol 1999;150: Sellers TA, Kushi LH, Potter JD, et al. Effect of family history, body-fat distribution, and reproductive factors on the risk of postmenopausal breast cancer. N Engl J Med 1992;326: Sellers TA, Gapstur SM, Potter JD, et al. Association of body fat distribution and family histories of breast and ovarian cancer with risk of postmenopausal breast cancer. Am J Epidemiol 1993;138: Potter JD, Cerhan JR, Sellers TA, et al. Progesterone and estrogen receptors and mammary neoplasia in the Iowa Women s Health Study: how many kinds of breast cancer are there? Cancer Epidemiol Biomarkers Prev 1995;4: Rebbeck TR, Walker AH, Phelan CM, et al. Defining etiologic heterogeneity in breast cancer using genetic biomarkers. Prog Clin Biol Res 1997;396: Enger SM, Ross RK, Paganini-Hill A, et al. Body size, physical activity, and breast cancer hormone receptor status: results from two case-control studies. Cancer Epidemiol Biomarkers Prev 2000;9: Bisgard KM, Folsom AR, Hong CP, et al. Mortality and cancer rates in nonrespondents to a prospective study of older women: 5-year follow-up. Am J Epidemiol 1994;139: Ries L, Kosary C, Hankey B, et al. SEER cancer statistics review, Bethesda, MD: National Cancer Institute, Lunn M, McNeil D. Applying Cox regression to competing risks. Biometrics 1995;51: Korn EL, Graubard BI, Midthune D. Time-to-event analysis of longitudinal follow-up of a survey: choice of the time-scale. Am J Epidemiol 1997;145: Soderqvist G. Effects of sex steroids on prolifen in normal mammary tissue. Ann Med 1998;30: Evans RM. The steroid and thyroid hormone receptor superfamily. Science 1988;240: Horwitz KB, Alexander PS. In situ photolinked nuclear progesterone receptors of human breast cancer cells: subunit molecular weights after transformation and translocation. Endocrinology 1983;113: Giangrande PH, Pollio G, McDonnell DP. Mapping and characterization of the functional domains responsible for the differential activity of the A and B isoforms of the human progesterone receptor. J Biol Chem 1997;272: Rowe SM, Coughlan SJ, McKenna NJ, et al. Ovarian carcinoma-associated TaqI restriction fragment length polymorphism in intron G of the progesterone receptor gene is due to an Alu sequence insertion. Cancer Res 1995;55: Wang-Gohrke S, Chang-Claude J, Becher H, et al. Progesterone receptor gene polymorphism is associated with decreased risk for breast cancer by age 50. Cancer Res 2000; 60: Manolitsas TP, Englefield P, Eccles DM, et al. No association of a 306-bp insertion polymorphism in the progesterone receptor gene with ovarian and breast cancer. Br J Cancer 1997; 75: Lancaster JM, Berchuck A, Carney ME, et al. Progesterone receptor gene polymorphism and risk for breast and ovarian cancer. (Letter). Br J Cancer 1998;78: Verhoog LC, Brekelmans CT, Seynaeve C, et al. Survival and tumour characteristics of breast-cancer patients with germline mutations of BRCA1. Lancet 1998;351: Johannsson OT, Idvall I, Anderson C, et al. Tumour biological features of BRCA1-induced breast and ovarian cancer. Eur J Cancer 1997;33: Thompson JA, Chen PL, King RA, et al. BRCA1 susceptibility markers and postmenopausal breast cancer: the Iowa Women s Health Study. Cancer Epidemiol Biomarkers Prev 2000;9: Thompson PA, Ambrosone C. Molecular epidemiology of genetic polymorphisms in estrogen metabolizing enzymes in human breast cancer. J Natl Cancer Inst Monogr 2000;(27):