Risk Factors for Breast Cancer in Elderly Women

American Journal Epidemiology Copyright 2004 by the Johns Hopkins Bloomberg School Public Health All rights reserved Vol. 160, 9 Printed in U.S.A. DOI: 10.1093/aje/kwh276 Risk Factors for Breast Cancer in Elderly Women Carol Sweeney 1,2, Cindy K. Blair 1, Kristin E. Anderson 1, DeAnn Lazovich 1, and Aaron R. Folsom 1 1 University Minnesota, Minneapolis, MN. 2 University Utah, Salt Lake City, UT. Received for publication December 31, 2003; accepted for publication May 11, 2004. As the population ages, an increasing fraction women diagnosed with breast cancer will be elderly. Heterogeneity breast cancer risk factors between pre- and postmenopausal women is recognized, but few studies have examined elderly women specifically. The authors describe the age-specific influence risk factors for postmenopausal breast cancer, with emphasis on women aged 75 or more years. Among 36,658 members the Iowa Women s Health Study (1986 2001), a population-based cohort study, 428 incident breast cancers were diagnosed in women aged 55 64 years, 1,297 in women aged 65 74 years, and 561 in women aged 75 84 years. Women with a body mass index (weight (kg)/height (m) 2 ) in the highest versus the lowest quartile were at increased risk breast cancer at age 75 or more years (adjusted hazard = 1.44, confidence interval (CI): 1.12, 1.84). Family history breast cancer (hazard = 1.54, CI: 1.24, 1.93 for a first-degree family history vs. none) and an older age at menopause (p trend = 0.07) conferred increased risk for women aged 75 or more years, and a high number livebirths was protective (hazard = 0.67, CI: 0.51, 0.88 for five or more births compared with one or two). Obesity, a modifiable risk factor, remained positively associated with breast cancer for all age groups postmenopausal women. age factors; breast neoplasms; female; risk factors Abbreviations: CI, confidence interval; IWHS, Iowa Women s Health Study. Increasing life expectancies in Western countries have resulted in an increase in the proportion the population that is elderly. For example, the percentage the US population that was aged 75 or more years increased from 2.6 percent in 1950 to 5.9 percent in 2000 (1). Because the incidence most cancers is highest at older ages, the proportion persons affected by cancer who are elderly at the time diagnosis is growing (2, 3). There is a need for research on factors influencing cancer etiology, prevention, and control in the elderly. The incidence breast cancer peaks at the ages 75 79 years, and among women diagnosed with breast cancer during 1996 2000, 44.2 percent were aged 65 or more years at diagnosis, and 22.5 percent were aged 75 or more years (4). Although some studies have reported on risk factors for breast cancer by age at diagnosis, and heterogeneity breast cancer risk factors between pre- and postmenopausal women is recognized, data describing risk factors for breast cancer in the oldest age groups are relatively limited. For example, in a report based on pooled data from multiple studies (5), only about 7 percent in the data set were aged 70 or more years at diagnosis. Among other studies that have considered the age-specific role breast cancer risk factors in postmenopausal women (6 14), most have treated all women aged 65 or more years as a single age group. Differences in the pathologic features breast tumors in aging women have been described. The proportion women with distant-stage disease increases with age (15, 16), the distribution histologic types changes (15), and the proportion tumors expressing hormone receptors (17) and other favorable biologic markers (18) increases. These trends appear to continue beyond age 70 years. Differences in pathology may reflect different biologic influences on breast tumor development in older women. We have analyzed data from a cohort in which a significant number women have been diagnosed with breast Correspondence to Dr. Carol Sweeney, Health Research Center, DFPM, 375 Chipeta Way, University Utah, Salt Lake City, UT 84108 (e-mail: Carol.Sweeney@hrc.utah.edu). 868

Breast Cancer in Elderly Women 869 cancer at age 75 or more years. The goal this analysis was to describe the age-specific influence recognized risk factors for breast cancer in aging women. MATERIALS AND METHODS Study population and follow-up The Iowa Women s Health Study (IWHS) was initiated in 1986 to examine the association between body fat and its distribution and cancer incidence in postmenopausal women. Subjects provided informed consent under a protocol approved for human subjects research by the University Minnesota Institutional Review Board. Details the methodology have been reported elsewhere (19, 20). Briefly, a questionnaire was mailed in January 1986 to a random sample women with an Iowa driver s license who were between the ages 55 and 69 years. The questionnaire solicited information regarding medical and reproductive histories, diet, and anthropometric, sociodemographic, and lifestyle factors. For the anthropometric variables, the participants reported their current weight and height, weight at age 18 years, and their hip and waist circumferences (measured, usually by a friend, using a paper tape measure provided with the questionnaire). The self-reported measurements in this cohort have been shown to be both valid and reliable (21). Forty-three percent women returned the questionnaire. Women were excluded who at baseline reported a history cancer other than skin cancer (n = 3,830), a prior mastectomy (n = 547), or premenopausal status (n = 354). Additionally, women who did not answer questions addressing the risk factors interest for our current analysis were excluded (n = 447). After exclusions, there were 36,658 women, 99 percent whom reported race as White, with a median age 61 years at baseline, in the eligible cohort. Incident breast cancer (International Classification Diseases code 174 and code C50 from the Ninth Revision and Tenth Revision, respectively) through 2001 were identified through annual linkage to the State Health Registry Iowa, which is part the National Cancer Institute s Surveillance, Epidemiology, and End Results Program (4). There was virtually no difference between the rates breast cancer among the responders and nonresponders to the baseline questionnaire (22). Person-years follow-up were calculated from the date the baseline questionnaire until the date breast cancer diagnosis, date move from Iowa, or date death. If none these events occurred, the end follow-up was December 31, 2001, when the median age 29,687 surviving participants was 77 years. Data analysis The primary exposures interest for the present analysis are recognized risk factors for postmenopausal breast cancer, including anthropometric measurements, reproductive events (age at first livebirth and parity), variables representing exposures to endogenous hormones (age at menarche, age at menopause), and family history breast cancer. From the self-reported anthropometric data, we calculated body mass index (weight (kg)/height (m) 2 ), waist/ hip (an indicator body fat distribution), and weight change from age 18 years to baseline, and we established cutpoints based on the distribution in all eligible cohort members. For family history breast cancer, we made comparisons 1) women who reported first-degree relatives (mothers, sisters, and daughters) diagnosed with breast cancer with those who reported no family history and 2) women who reported second-degree relatives diagnosed with breast cancer (grandmothers and aunts), but no firstdegree relatives, with the same referent group. To analyze associations between risk factors and incident breast cancer, we used the Cox proportional hazards regression method with age as the time scale. To control for potential confounders the variables interest, we selected covariates considered to be risk factors according to previous breast cancer studies, including previous IWHS analyses. Multivariate regression analyses were conducted with the following covariates: age at baseline (continuous), education (less than high school, high school, more than high school), age in years at first livebirth (<20, 20 24, 25 29, 30, nulliparous), parity (1 2, 3 4, 5, nulliparous), age in years at menarche ( 11, 12, 13, 14), age in years at menopause (<45, 45 49, 50 54, 55), family history breast cancer (none, second-degree relative but no first, first-degree relatives), and body mass index and height (quartiles). Hazard s and 95 percent confidence intervals were calculated for each three age intervals: 55 64 years, 65 74 years, and 75 84 years. Each woman contributed person-years at risk to one or more these age categories, depending on her age at entry into the study and her age at end follow-up. Trends in hazard s across ordered exposure categories (e.g., quartiles body mass index) were examined using a continuous variable (e.g., coded 0, 1, 2, 3) as the explanatory variable. Because the cutpoints for 10-year categories may be arbitrary, we constructed graphs hazard s across age. For these graphs, we calculated a point estimate the hazard for breast cancer at each single year age from Cox models that included time-dependent covariates. The point estimates were plotted versus age on a log scale. A line representing the trend in hazard with age was obtained by linear regression the single-year estimates across age, weighting each estimate according to the number incident breast cancers at that age. We formally assessed differences in hazard s across the three categories age at diagnosis by comparing two models. The first model included time-dependent covariates for the risk factor interest, allowing a different hazard to be estimated for each three age intervals: 55 64, 65 74, and 75 84 years. The second model did not include time-dependent covariates, constraining the hazard to be constant across all age intervals. Interaction between risk factors and age was tested by comparing the two models using the likelihood test. Stata statistical stware (Stata Corpon, College Station, Texas) was used for data analysis. RESULTS During 16 years follow-up, there were 2,286 women aged 56 84 years who were identified with incident breast

870 Sweeney et al. TABLE 1. Characteristics breast cancer by age, Iowa Women s Health Study, 1986 2001 Age at diagnosis 55 59 years 60 64 years 65 69 years 70 74 years 75 79 years 80 84 years % % % % % % Stage In situ 3 4.4 47 13.1 75 11.5 91 14.1 66 15.2 12 9.5 Local 38 55.9 202 56.1 414 63.6 411 63.6 260 59.8 85 67.5 Regional 24 35.3 93 25.8 139 21.4 107 16.6 71 16.3 19 15.1 Distant 2 2.9 13 3.6 17 2.6 20 3.1 14 3.2 3 2.4 Unstaged 1 1.5 5 1.4 6 0.9 17 2.6 24 5.5 7 5.6 Estrogen receptor status Positive 40 58.8 207 57.5 419 64.4 417 64.6 289 66.4 86 68.3 Negative 9 13.2 42 11.7 90 13.8 80 12.4 43 9.9 8 6.4 Unknown 19 27.9 111 30.8 142 21.8 149 23.1 103 23.7 32 25.4 Histology Lobular 2 2.9 24 6.7 47 7.2 52 8.6 35 8.9 7 7.7 Ductal 62 91.2 316 87.8 563 86.5 528 86.8 332 84.5 78 85.7 Other 4 5.9 20 5.6 41 6.3 28 4.6 26 6.6 6 6.6 Person-years at risk 30,911 93,289 143,295 139,451 76,507 24,810 incident 68 360 651 646 435 126 Rate/100,000 person-years 220 386 454 463 569 508 cancer. Because the age structure the cohort, restricted to ages 55 69 years in 1986, there were many breast cancer available for analysis at ages 65 74 years, but there were fewer at the youngest and oldest ages. The maximum number breast cancers diagnosed at a single year age was 157 at age 69 years; the minimum was nine at age 56 years. Characteristics breast cancer by age at diagnosis are shown in table 1. Breast cancer incidence was highest in women aged 75 79 years. The majority tumors were local stage at diagnosis (63 percent tumors with stage information), with the proportion that was local generally increasing with age. The proportion tumors positive for the estrogen receptor (84 percent all those with estrogen receptor results) also increased with age. The proportion with ductal histology decreased with age at diagnosis, while the proportion with histologies classified as other (including mucinous adenocarcinoma, tubular adenocarcinoma, medullary carcinoma) was highest in women aged 75 or more years. Hazard s for anthropometric risk factors for breast cancer diagnosed at the ages 55 64, 65 74, and 75 84 years are presented in table 2. A trend increasing risk associated with each quartile increase in body mass index above the lowest was observed, and this trend was apparent in every age group. When the hazard s per 4-kg/m 2 increase in body mass index at each year age at diagnosis were plotted (figure 1), there was no indication that the increased risk associated with body mass index changed with age, and the fitted line hazard s across age had essentially zero slope. Other measures body fatness, that is, waist/hip and weight change since age 18 years, also showed consistent trends increasing risk within each age group (table 2). Risk breast cancer increased with height in all age groups, although the trend was not significant for women aged 55 64 years. Age-specific hazard s for associations between reproductive history and breast cancer are shown in table 3. There was no increased risk breast cancer diagnosis at age 75 or more years associated with an older age at first birth. However, in this cohort the magnitude the increased risk for older age at first birth was also modest for the other ageat-diagnosis categories after adjustment for number livebirths and education. Nulliparous women were at increased risk breast cancer relative to women with a first birth before age 20 years, but hazard s were attenuated in the older age groups. The inverse association parity five or more with breast cancer was observed for all three age categories. The hazard s for nulliparous women and for women with five or more livebirths, calculated at each single year age, are shown in figure 2, parts A and B. The slopes the hazard s trended toward the null at older ages; however, there was no statistical evidence an interaction between nulliparity (p = 0.27) or parity five or more (p = 0.49) and age at diagnosis.

Breast Cancer in Elderly Women 871 TABLE 2. Anthropometric risk factors for breast cancer in postmenopausal women by age at diagnosis, Iowa Women s Health Study, 1986 2001 Age at diagnosis 55 64 years 65 74 years 75 84 years Rate* Hazard CI * Incidence per 100,000 person-years. CI, confidence interval. Adjusted for age at baseline, education, age at first livebirth, parity, age at menarche, age at menopause, family history breast cancer, and height. Number does not sum to 2,286 because some subjects did not report waist, hip, or weight at age 18 years. One pound = 0.45 kg. # Adjusted for age at baseline, education, age at first livebirth, parity, age at menarche, age at menopause, family history breast cancer, and body mass index. ** One inch = 2.54 cm. Rate Hazard CI Rate Body mass index (kg/m 2 ) <23.5 101 32,529 310 1 Referent 274 71,506 383 1 Referent 112 24,886 450 1 Referent 23.5 26 78 29,500 264 0.86 0.64, 1.16 306 67,332 454 1.21 1.03, 1.42 129 24,590 525 1.19 0.92, 1.53 >26 29.5 119 30,724 387 1.26 0.96, 1.64 335 72,159 464 1.26 1.08, 1.49 167 26,532 629 1.45 1.14, 1.85 >29.5 130 31,449 413 1.34 1.03, 1.75 382 71,752 532 1.48 1.26, 1.73 153 25,311 604 1.44 1.12, 1.84 p trend 0.004 <0.0001 0.001 Waist/hip, <0.78 117 38,141 307 1 Referent 312 74,694 418 1 Referent 102 22,966 444 1 Referent 0.78 <0.83 91 29,231 311 1.02 0.77, 1.34 296 65,598 451 1.10 0.94, 1.30 128 23,239 551 1.27 0.98, 1.64 0.83 <0.89 102 29,264 349 1.14 0.87, 1.48 308 70,196 439 1.10 0.93, 1.28 149 26,446 563 1.31 1.02, 1.69 0.89 115 27,038 425 1.38 1.06, 1.80 375 71,103 527 1.34 1.15, 1.56 179 28,342 632 1.49 1.16, 1.90 p trend 0.01 0.0004 0.002 Weight change, age 18 to baseline (pounds),, Loss or gain, 13 102 33,353 306 1 Referent 252 73,607 342 1 Referent 100 25,500 392 1 Referent Gain, 14 28 89 31,580 282 0.93 0.70, 1.24 318 69,791 456 1.34 1.14, 1.58 136 24,790 549 1.39 1.07, 1.80 Gain, 29 45 100 29,139 343 1.13 0.85, 1.48 308 69,224 445 1.31 1.11, 1.55 148 26,202 565 1.42 1.10, 1.84 Gain, >45 133 29,705 448 1.47 1.13, 1.91 413 68,900 599 1.78 1.52, 2.08 175 24,274 721 1.79 1.40, 2.30 p trend 0.001 <0.0001 <0.0001 Height (inches)#,** 62 103 30,950 333 1 Referent 291 76,462 381 1 Referent 142 29,423 483 1 Referent 63 64 129 37,195 347 1.06 0.82, 1.37 399 88,691 450 1.18 1.02, 1.38 179 33,739 531 1.11 0.89, 1.38 65 66 117 33,632 348 1.09 0.84, 1.43 365 72,141 506 1.34 1.15, 1.57 144 23,689 608 1.27 1.01, 1.61 >66 79 22,425 352 1.11 0.82, 1.49 242 45,455 532 1.41 1.19, 1.67 96 14,467 664 1.40 1.08, 1.82 p trend 0.46 <0.0001 0.005 Hazard CI Incidence rates for breast cancer increased with age in all categories parous women (table 3), but the incidence rates among nulliparous women did not fit this pattern. The incidence breast cancer among nulliparous women increased from 447 per 100,000 person-years (95 percent confidence interval (CI): 321, 607) at age 55 64 years to 496 per 100,000 (95 percent CI: 412, 591) at age 65 74 years but then decreased slightly to 486 per 100,000 (95 percent CI: 364, 636) at age 75 84 years. Only 9 percent the women in the IWHS cohort reported having had no livebirths, and therefore the confidence intervals on the incidence rates for breast cancer for nulliparous women were wide. Menarche at older than 11 years was associated with a decreased risk breast cancer at age 55 64 years (table 3), although there was not a clear trend across the categories women who reported menarche at the ages 12, 13, and 14 or more years. The hazard s for older age at menarche were attenuated for breast cancer diagnosed at ages 65 74 and 75 84 years compared with age 55 64 years. Older age at menopause was modestly associated with increased risk breast cancer. When the contrast was simplified to compare two categories age at menopause, hazard s for menopause at age 50 or more years, compared with that at age 49 or less years, were 1.11 (95 percent CI: 0.92, 1.35) for age 55 64 years at diagnosis, 1.05 (95 percent CI: 0.94, 1.17) for age 65 74 years at diagnosis, and 1.26 (95 percent CI: 1.06, 1.49) for age 75 84 years at diagnosis. The graph hazard s for age at menopause 50 or more years versus less

872 Sweeney et al. FIGURE 1. Breast cancer risk associated with body mass index among elderly women, by age at diagnosis, Iowa Women s Health Study, 1986 2001. Each circle indicates the age-specific hazard per 4-kg/m 2 increment in body mass index; the diameter the circle is proportional to the number breast cancers diagnosed at that age. The dashed line is a weighted linear fit the individual hazard s. than 50 years (figure 2, part C) has a slight positive slope, but the test for interaction did not indicate an important difference in the hazard for age at menopause by age at diagnosis (p = 0.60). In the present data, there was little relation between family history breast cancer and breast cancer risk at age 55 64 years, but increased risk associated with family history was present for ages 65 74 years and 75 84 years. The graph hazard s for first-degree family history breast cancer at each year age had a positive slope with age (figure 2, part D). The test for interaction provided limited evidence (p = 0.10) heterogeneity the hazard s for a firstdegree family history breast cancer among the three age groups. DISCUSSION As the population ages, elderly women will represent an increasing fraction women diagnosed with breast cancer (2, 3). Using data from a large, population-based cohort, we found that parity, age at menopause, and family history continued to influence breast cancer risk among women aged 75 or more years. High body mass index, a modifiable risk factor, was consistently associated with increased breast cancer risk for all age groups in this cohort postmenopausal women. Breast cancer incidence in this cohort followed the age trend reported in national data (4), with increasing incidence in each 5-year age group up to 75 79 years and then somewhat lower incidence at age 80 84 years. Overall similarity breast cancer incidence between IWHS participants and the population from which they were drawn has been described previously (22). Breast cancers occurring in women who were aged 75 or more years at diagnosis were characterized by a somewhat higher proportion with three favorable prognostic characteristics: local stage at diagnosis, positive estrogen receptor status, and other histologies (tumors that were neither ductal nor lobular). The patterns for estrogen receptor status and histology are in agreement with what has previously been described (15, 17, 18). In Surveillance, Epidemiology, and End Results Program data from the 1980s (15) and from 1992 (16), women in the oldest age groups more ten had advanced stage breast cancer. This was not true in the more recently diagnosed IWHS, a difference that is probably explained by the continued increase in use mammography screening during the last decade by women aged 65 or more years (23). The strengths this study were that it was population based and prospective, with the same participants being followed continuously over a period 16 years for breast cancer incidence. Because the cohort was followed from a median age 61 years to a median age 77 years, most cohort members contributed person-time in all three age categories that were considered in the analysis. To further control for any possible birth cohort effects, analyses were adjusted for age at baseline. Generalizability is a potential limitation; IWHS participants live in one state, are almost exclusively Caucasian, had to have a driver s license in 1986, and had to be able to respond to a detailed written questionnaire. Thus, these women differ from the general population US women at risk for breast cancer in racial and geographic distribution and literacy. For this analysis, we have emphasized comparisons the disease experience women within the cohort at different ages. The results therefore should be internally valid. The time between measurement exposures at baseline and occurrence breast cancer varied widely. When baseline responses are used to assess hazard s for disease events through a long follow-up period, accuracy exposure information may diminish with time, for example, if subjects change their dietary habits from what was reported at baseline. There is the potential for hazard s for events occurring later during follow-up to be attenuated because misclassification exposure. For most the factors that we emphasized in this analysis, however, for example, parity, age at menarche, and age at menopause, the exposure could not change after baseline among these postmenopausal women, so any attenuation hazard s with age would not be explained by this source measurement error. In the existing literature on breast cancer risk factors, age heterogeneity is most evident when the role body fatness, usually measured by body mass index, is considered. Many studies have shown that the relation between high body mass index and breast cancer is null or slightly inverse when cancers diagnosed before menopause are considered, but high body mass index is associated with an increased risk breast cancer for postmenopausal women (5, 11, 24). Among prior studies that considered the role high body mass index by age group among postmenopausal women, most have suggested that relative risks were stronger in the oldest age groups (5, 11 13), although one reported the reverse (25). For example, in one study the odds for the highest versus the lowest quintile body mass index increased from 1.5 for women diagnosed at age 60 69 years to 2.9 among women aged 70 or more years at diagnosis (11). Those analyses were based on relatively small numbers in the oldest age

Breast Cancer in Elderly Women 873 TABLE 3. Reproductive and family history risk factors for breast cancer in postmenopausal women by age at diagnosis, Iowa Women s Health Study, 1986 2001 Age at diagnosis 55 64 years 65 74 years 75 84 years Rate* Hazard CI * Incidence per 100,000 person-years. CI, confidence interval. Adjusted for age at baseline, education, body mass index, height, and other exposures in this table. Number does not sum to 2,286 because some subjects did not report age at menarche, age at menopause, or family history breast cancer. Rate Hazard CI Rate Hazard Age at first livebirth (years) <20 92 29,562 311 1 Referent 219 54,936 399 1 Referent 83 16,182 513 1 Referent 20 24 209 61,766 338 1.02 0.79, 1.32 576 130,394 442 1.07 0.91, 1.26 231 41,468 557 1.02 0.79, 1.32 25 29 66 18,999 347 0.99 0.71, 1.39 283 56,428 502 1.22 1.01, 1.47 144 24,649 584 1.01 0.75, 1.35 30 20 4,713 424 1.15 0.69, 1.91 95 15,995 594 1.42 1.10, 1.84 50 8,116 616 1.00 0.69, 1.45 Nulliparous 41 9,163 447 1.42 0.96, 2.11 124 24,996 496 1.24 0.97, 1.58 53 10,903 486 1.13 0.77, 1.65 p trend 0.79 0.003 0.97 livebirths 1 2 129 35,010 368 1 Referent 429 90,334 475 1 Referent 226 36,225 624 1 Referent 3 4 189 52,595 359 0.99 0.78, 1.24 526 112,227 469 1.00 0.87, 1.14 208 37,375 557 0.86 0.71, 1.04 5 69 27,434 252 0.69 0.51, 0.93 218 55,192 395 0.86 0.72, 1.02 74 16,816 440 0.67 0.51, 0.88 p trend 0.02 0.12 0.004 Age at menarche (years), 11 90 21,162 425 1 Referent 201 42,720 471 1 Referent 78 13,507 577 1 Referent 12 110 33,033 333 0.79 0.60, 1.05 383 75,699 506 1.08 0.91, 1.29 134 27,037 496 0.86 0.65, 1.14 13 116 36,304 320 0.79 0.59, 1.04 374 83,175 450 0.98 0.82, 1.16 175 30,423 575 1.02 0.78, 1.33 14 109 32,640 334 0.82 0.61, 1.09 330 78,152 422 0.93 0.77, 1.11 171 29,049 589 1.03 0.79, 1.36 p trend 0.24 0.13 0.34 Age at menopause (years), <45 109 29,873 365 1 Referent 289 65,045 444 1 Referent 122 22,661 538 1 Referent 45 49 97 32,790 296 0.84 0.64, 1.10 320 71,120 450 1.00 0.86, 1.18 107 24,849 431 0.80 0.62, 1.04 50 54 167 49,023 341 0.95 0.75, 1.21 510 106,944 477 1.06 0.91, 1.22 224 36,927 607 1.11 0.89, 1.39 55 48 9,228 520 1.36 0.97, 1.92 141 30,055 469 1.04 0.85, 1.28 81 12,818 632 1.16 0.87, 1.53 p trend 0.26 0.45 0.07 Family history breast cancer, None 302 92,094 328 1 Referent 861 207,316 415 1 Referent 377 73,467 513 1 Referent Second degree 54 14,772 366 1.11 0.83, 1.48 161 30,411 529 1.25 1.05, 1.48 62 9,990 621 1.19 0.91, 1.56 First degree 51 13,241 385 1.15 0.86, 1.55 223 33,058 675 1.61 1.39, 1.87 100 12,530 798 1.54 1.24, 1.93 p trend 0.28 <0.0001 0.0002 CI groups. In the IWHS cohort, based on 1,297 diagnosed at age 65 74 years and 561 diagnosed at age 75 or more years, we observed that the magnitude the hazard for body mass index in postmenopausal breast cancer was similar across all ages at diagnosis. Several measures body fatness (body mass index, waist/hip, and weight change since age 18 years) each had a strong positive association with breast cancer risk for women in all three age groups that we considered. Thus, it appears that women who heed public health messages to maintain a healthy weight in adulthood will have lower breast cancer risk in their elderly years. An early age at first livebirth and a higher number births are recognized as protective against postmenopausal breast cancer (26). The present data provide evidence that, among parous women, a high number births ( 5 births) is associated with decreased breast cancer risk in all age groups postmenopausal women, including those aged 75 or more years. We found that nulliparity and age at first livebirth had little association with breast cancer in those aged 75 or more years. Several other studies that reported on parity, age at first birth, and breast cancer by age at diagnosis (6 9) found higher risk s overall for nulliparous women than did the

874 Sweeney et al. FIGURE 2. Breast cancer risk associated with reproductive history or family history breast cancer among elderly women, by age at diagnosis, Iowa Women s Health Study, 1986 2001. A, hazard for nulliparity (referent group is age at first birth <20 years); B, hazard for five or more livebirths (referent group is 1 4 births); C, hazard for age at menopause 50 or more years (referent group is age at menopause less than 50 years); D, hazard for first-degree family history breast cancer (referent group is women with no family history). Each circle indicates the age-specific hazard ; the diameter the circle is proportional to the number breast cancers diagnosed at that age. The dashed line is a weighted linear fit the individual hazard s. present study, and these studies did not find diminished risk s among elderly women. The inconsistency between the result for women aged 75 or more years in the present study and these reports may be explained by chance. There is also the possibility that categorizing women aged 75 or more years as a separate group, rather than grouping all women aged 65 or more years (7 9) or 70 or more years (6), may be revealing different information. Older age at menarche typically is reported to be associated with reduced breast cancer risk, while older age at menopause is associated with increased risk (26). In our analysis, the association with age at menarche was present only in the group aged 55 64 years, with little evidence an effect age at menarche on breast cancer risk in women aged 65 or more years. In the present data, the association between older age at menopause and breast cancer risk was not strong but was present in women aged 75 or more years. These results are consistent with an analysis based on case-control study data, which also indicated that age at menarche was no longer relevant to breast cancer risk after the age 64 years, but age at menopause continued to influence risk (10). A pooled analysis data from population-based studies reported on the association between family history breast cancer and breast cancer risk by 5-year age categories (27). The risk for having one affected relative compared with none decreased from 2.53 among women aged 35 39 years, to 1.84 for women aged 45 49 years, to 1.53 for women aged 55 59 years, consistent with the idea that inheritance a high-risk cancer susceptibility gene is associated with both an increased risk cancer and a tendency toward earlier age at onset. In the same study, risk s did not further diminish in older age groups; for women aged 70 or more years, the risk was 1.64. Among the postmenopausal women in the IWHS, the hazard for a first-degree family history breast cancer was higher in the groups aged 65 74 and 75 84 years than in the group aged 55 64 years. The results the present study within a single cohort observed over 16 years are in agreement with the pooled analysis multiple studies (27)

Breast Cancer in Elderly Women 875 in indicating that family history breast cancer, and by implication genetic susceptibility, continues to predict elevated breast cancer risk in elderly women. Epidemiologic theory predicts that risk s will be stronger in populations with lower absolute risk (28). An example this phenomenon can be drawn from cardiovascular disease, for which recognized risk factors, for example, smoking or high blood pressure, show larger risk s for younger, low-risk subjects than for older, high-risk subjects within the same cohorts (29). The findings lack association nulliparity, age at first livebirth, and age at menarche with breast cancer at age 75 or more years in the IWHS are consistent with attenuation risk s as breast cancer incidence rises with age. However, for several other breast cancer risk factors, we did not observe lower risk s for women aged 75 or more years. The risk factors that showed attenuated risk s, that is, age at menarche, nulliparity, and age at first birth, represent hormonal exposures that occurred in the distant past. Indicators relatively recent exposure to endogenous hormones (postmenopausal obesity and age at menopause) appear to remain relevant for breast cancer risk in elderly women. ACKNOWLEDGMENTS This research was supported by National Cancer Institute grant R01 CA39742. The authors thank Ching-Ping Hong, Dr. David R. Jacobs, and Peter Hannan for useful suggestions regarding the data analysis. REFERENCES 1. Hobbs F, Stoops N. Demographic trends in the 20th century. Washington, DC: US Government Printing Office, 2002. 2. Edwards BK, Howe HL, Ries LA, et al. Annual report to the nation on the status cancer, 1973 1999, featuring implications age and aging on U.S. cancer burden. Cancer 2002;94:2766 92. 3. Alberg AJ, Singh S. Epidemiology breast cancer in older women: implications for future healthcare. Drugs Aging 2001; 18:761 72. 4. Ries LAG, Eisner MP, Kosary CL, et al. SEER cancer statistics review, 1975 2000. Bethesda, MD: National Cancer Institute, 2003. 5. van den Brandt PA, Spiegelman D, Yaun SS, et al. Pooled analysis prospective cohort studies on height, weight, and breast cancer risk. Am J Epidemiol 2000;152:514 27. 6. Lubin JH, Burns PE, Blot WJ, et al. Risk factors for breast cancer in women in northern Alberta, Canada, as related to age at diagnosis. J Natl Cancer Inst 1982;68:211 17. 7. Kvale G, Heuch I, Eide GE. A prospective study reproductive factors and breast cancer. I. Parity. Am J Epidemiol 1987;126: 831 41. 8. Bouchardy C, Le MG, Hill C. Risk factors for breast cancer according to age at diagnosis in a French case-control study. J Clin Epidemiol 1990;43:267 75. 9. Lund E. Breast cancer mortality and the change in fertility risk factors at menopause: a prospective study 800,000 married Norwegian women. Epidemiology 1991;2:285 8. 10. La Vecchia C, Negri E, Bruzzi P, et al. The role age at menarche and at menopause on breast cancer risk: combined evidence from four case-control studies. Ann Oncol 1992;3:625 9. 11. Franceschi S, Favero A, La Vecchia C, et al. Body size indices and breast cancer risk before and after menopause. Int J Cancer 1996;67:181 6. 12. Yong LC, Brown CC, Schatzkin A, et al. Prospective study relative weight and risk breast cancer: the Breast Cancer Detection Demonstn Project follow-up study, 1979 to 1987 1989. Am J Epidemiol 1996;143:985 95. 13. Galanis DJ, Kolonel LN, Lee J, et al. Anthropometric predictors breast cancer incidence and survival in a multi-ethnic cohort female residents Hawaii, United States. Cancer Causes Control 1998;9:217 24. 14. Li CI, Malone KE, Porter PL, et al. Reproductive and anthropometric factors in relation to the risk lobular and ductal breast carcinoma among women 65 79 years age. Int J Cancer 2003; 107:647 51. 15. Yancik R, Ries LG, Yates JW. Breast cancer in aging women. A population-based study contrasts in stage, surgery, and survival. Cancer 1989;63:976 81. 16. Yancik R, Wesley MN, Ries LA, et al. Effect age and comorbidity in postmenopausal breast cancer patients aged 55 years and older. JAMA 2001;285:885 92. 17. Li CI, Daling JR, Malone KE. Incidence invasive breast cancer by hormone receptor status from 1992 to 1998. J Clin Oncol 2003;21:28 34. 18. Diab SG, Elledge RM, Clark GM. Tumor characteristics and clinical outcome elderly women with breast cancer. J Natl Cancer Inst 2000;92:550 6. 19. Folsom AR, Kaye SA, Prineas RJ, et al. Increased incidence carcinoma the breast associated with abdominal adiposity in postmenopausal women. Am J Epidemiol 1990;131:794 803. 20. Folsom AR, Kushi LH, Anderson KE, et al. Associations general and abdominal obesity with multiple health outcomes in older women: the Iowa Women s Health Study. Arch Intern Med 2000;160:2117 28. 21. Kushi LH, Kaye SA, Folsom AR, et al. Accuracy and reliability self-measurement body girths. Am J Epidemiol 1988;128: 740 8. 22. Bisgard KM, Folsom AR, Hong C, et al. Mortality and cancer rates in nonrespondents to a prospective study older women: 5- year follow-up. Am J Epidemiol 1994;139:990 1,000. 23. Centers for Disease Control and Prevention. Behavioral Risk Factor Surveillance System survey data. Atlanta, GA: Centers for Disease Control and Prevention, US Department Health and Human Services, 2004. (http://apps.nccd.cdc.gov/brfss/trends/ TrendData.asp). 24. Le Marchand L, Kolonel LN, Earle ME, et al. Body size at different periods life and breast cancer risk. Am J Epidemiol 1988; 128:137 52. 25. Morimoto LM, White E, Chen Z, et al. Obesity, body size, and risk postmenopausal breast cancer: the Women s Health Initiative (United States). Cancer Causes Control 2002;13:741 51. 26. Kelsey JL, Gammon MD, John EM. Reproductive factors and breast cancer. Epidemiol Rev 1993;15:36 47. 27. Collaborative Group on Hormonal Factors in Breast Cancer. Familial breast cancer: collaborative reanalysis individual data from 52 epidemiological studies including 58,209 women with breast cancer and 101,986 women without the disease. Lancet 2001;358:1389 99. 28. Rothman KJ, Poole C. A strengthening programme for weak associations. Int J Epidemiol 1988;17:955 9. 29. Psaty BM, Koepsell TD, Manolio TA, et al. Risk s and risk differences in estimating the effect risk factors for cardiovascular disease in the elderly. J Clin Epidemiol 1990;43:961 70.