Findings on increased risk in women with a family history of breast cancer have been more consistent. A meta-analysis [3] of five studies that had examined the issue suggested that the risk for breast cancer associated with HRT was elevated among women without a family history of breast cancer (relative risk [RR], 1.5 [95% CI, 1.2 to 1.7]) but was even greater among women with a family history (RR, 3.4 [CI, 2.0 to 6.0]). Of the studies not included in the meta-analysis, some [9-12] but not all [6, 7, 13] reported similar observations.

Beyond the issue of the effects of HRT on the incidence of breast cancer, further consideration of the concomitant effect of HRT on risk for death is warranted. A recent report from the Nurses' Health Study [8] suggests that hormonal therapy lasting at least 5 years is associated with a 45% increase in the rate of death from breast cancer. Analytic epidemiologic studies suggest that women who receive HRT have a significantly lower risk for death, primarily because of protection against cardiovascular disease [14-17]. Thus, it is important to examine all-cause mortality when considering the risks and benefits of HRT use [18], especially among women with a family history of breast cancer.

As part of the Iowa Women's Health Study, a prospective cohort study of 41 837 women who were 55 to 69 years of age at baseline in 1986, data were collected on self-reported family history of breast cancer in first-degree relatives and on history of HRT use. These data were used to examine the risks for postmenopausal breast cancer, case fatality, and total mortality.

Methods

Top Methods Results Discussion Author & Article Info References

The Iowa Women's Health Study Cohort

Participants in the Iowa Women's Health Study were selected from a random sample of all women between 55 and 69 years of age who had a valid Iowa driver's license in 1985 [19]. This sampling frame comprised approximately 94% of female Iowa residents in that age range. The primary purpose of the study was to examine the association of obesity and body fat distribution with the incidence of cancer and mortality rate. A total of 41 837 women returned a mailed questionnaire (response rate, 42.7%). Three follow-up questionnaires mailed in October 1987, August 1989, and June 1992 were used to collect information on incident self-reported disease and each participant's current residence. Only minor demographic differences were seen at baseline between respondents and nonrespondents [20]. Nonrespondents have subsequently been found to have a higher 5-year incidence of cancer and rate of death (primarily death from diseases related to smoking) [21]. The incidence of breast cancer, however, was similar among respondents and nonrespondents (RR, 1.01 [CI, 0.9 to 1.1]).

Self-reported items on the baseline questionnaire included education level; marital status; smoking history; usual alcohol intake during the past year; physical activity; reproductive history; and history of various medical conditions, including diabetes and heart disease. We asked women whether they had menstruated in the past year and asked them to state the age at which and the reason why their periods had stopped (natural menopause, surgery, medication, or other). We asked them whether they had "ever used pills, other than birth control pills, which contain estrogen or other female hormones" and how long they had used them. Response categories for duration of use were "one month or less," "2-6 months," "7-12 months," "13 months-2 years," "3-5 years," or "more than 5 years." Participants were asked whether their mothers, sisters, or daughters had ever had breast cancer. No information was collected on the number of relatives in a given category or on the age of relatives. For exclusion purposes, we ascertained the presence of prevalent cancer by asking participants whether they had ever been told by a physician that they had any form of cancer, excluding skin cancer. Participants reported their current height and weight and their weight at 18 years of age. A paper measuring tape was sent so that a friend, spouse, or relative could measure the circumferences of the waist (1 inch above the umbilicus) and the hips (maximum). The height, weight, and circumference measurements obtained by this protocol are valid and reliable [22]. Anthropometric values were used to derive current body mass index (weight in kg/height in m²), body mass index at 18 years of age (based on their current height), and the ratio of waist circumference to hip circumference. The second follow-up survey included questions on use of mammography, including whether the participants had ever undergone mammography and the length of time since the last mammography had been done.

Cohort Follow-up

Cases of breast cancer (codes 500 to 509 of the International Classification of Diseases for Oncology, Second Edition) among Iowa residents were ascertained through the State Health Registry of Iowa, which is part of the National Cancer Institute's Surveillance, Epidemiology, and End Results Program (SEER) [23]. Each year, registry cases and cohort members were matched by a computer according to social security number, name, maiden name, and date of birth. Primary tumor site, date of diagnosis, nonspecific extent of disease, and tumor size were obtained for each incident cancer case from 1986 through 1993.

Information on deaths was obtained from the State Health Registry of Iowa for Iowa residents and from the National Death Index for women who did not respond to the three follow-up surveys. Causes of death were coded according to the International Classification of Diseases, Ninth Revision [24].

Data Analysis

Participants with the following baseline characteristics were excluded from the analysis: premenopausal status (n = 569), total or partial mastectomy (n = 1870), history of any cancer other than skin cancer (n = 2293), and unknown family history of breast cancer (n = 1186). After women with these characteristics were excluded, 35 919 women remained in the at-risk cohort for analysis.

Women were considered to be at risk for breast cancer from January 1986 through 31 December 1993 or until one of the following censoring events occurred: diagnosis of breast cancer, death, relocation to an area other than Iowa, or other loss to follow-up. Person-years of follow-up for the mortality analyses were calculated from completion of the 1986 baseline questionnaire until death or the end of 1993.

Characteristics of the women were collapsed into natural categories or quartiles. Use of HRT was categorized by status ("current," "former," and "never") at baseline and by duration ( 5 years or >5 years).

To describe the association of family history, HRT, and other variables with the incidence of breast cancer and total mortality rate, we computed age-adjusted rates and 95% CIs [25]; multivariate-adjusted relative risks and 95% CIs were derived through proportional hazards regression models by using the SAS program PHREG (SAS Institute, Cary, North Carolina) [26]. Multiplicative interactions were evaluated by comparing the –2 log likelihoods of models with and without the cross-product (interaction) terms within the context of proportional hazards regression.

Results

Top Methods Results Discussion Author & Article Info References

Risk Factors and Incidence of Breast Cancer

After 8 years and more than 275 000 person-years of observation, 1085 cases of postmenopausal breast cancer were documented. The association of accepted risk factors [27] with incidence of breast cancer was evident in this cohort. Early age at menarche, late age at first birth, high waist-to-hip ratio, high body mass index, education, and alcohol intake were associated with increased risk; high body mass index at 18 years of age was associated with decreased risk (Table 1). All subsequent analyses of the incidence of breast cancer include adjustment for these risk factors in addition to current age, type of menopause, and age at menopause.

Hormone Replacement Therapy and Incidence of Breast Cancer by Family History

A family history of breast cancer in mothers or sisters was reported by 12.2% of the cohort at risk. Approximately 38% of the cohort at risk reported having ever received HRT. The frequency of reported use did not differ by family history: 38.3% of women without a family history and 37.7% of women with a family history (P > 0.2). Duration of use was similar (P > 0.2).

After adjustment for the risk factors listed in Table 1, the relative risks for breast cancer among women who had formerly received HRT for 5 years or less or for more than 5 years were 1.04 and 0.89, respectively. The multivariate adjusted relative risks for current short-term ( 5 years) and long-term (>5 years) users of HRT were 1.34 (CI, 0.98 to 1.82) and 1.17 (CI, 0.90 to 1.51), respectively. The age-adjusted incidence rates of breast cancer associated with HRT use by family history of breast cancer are shown in Table 2. Among women without a family history of breast cancer, the rate of breast cancer in those who were currently receiving HRT and had been receiving it for at least 5 years was 41 cases per 10 000 person-years, a rate not significantly greater than the 36 per 10 000 person-years seen in women who had never received HRT. This translates into a rate difference of 5 per 10 000 person-years. Multivariate adjustment for other risk factors for breast cancer did not change these results; the relative risk in women who were currently receiving therapy and had been receiving it for at least 5 years was 1.13 (not statistically significant). Rates of breast cancer among women with a family history of breast cancer were 15 per 10 000 person-years greater (CI, –19 to 50 per 10 000 person-years) for current long-term users of HRT (61 per 10 000 person-years compared with 46 per 10 000 person-years in women who had never received HRT). The relative risks were higher in women who currently received HRT than in those who formerly received HRT (regardless of duration of use), but none of the relative risks was statistically significant. Results of a formal test for interaction between family history and HRT use were not statistically significant (P > 0.2).

Family History, Mammography, and Incidence of Breast Cancer

One possible explanation for the slightly increased risk associated with HRT use among women with a family history of breast cancer is the increased surveillance provided by screening mammography. Among the 35 919 women at risk for breast cancer, data on mammography were self-reported by 31 691 women who responded to the second follow-up survey in 1989. Participation rates on the follow-up survey by women with a family history of breast cancer did not differ from the rates at baseline (12.2% at both surveys). Sixty-five percent reported that they had ever undergone mammography: 37% within 1 year of the survey and 18.8% within 2 years. Women with a family history of breast cancer were more likely than women without a family history to have ever had mammography (75.4% compared with 63.6%; P < 0.001) and were more likely to have had mammography in the 2 years before the survey (P < 0.001). Moreover, a history of mammography was positively associated with HRT: Use of HRT was reported by 44% of the women who had previously had mammography but by only 30% of women who had never had mammography (P < 0.001). Despite the apparent increased surveillance for women reporting HRT use and a family history of breast cancer, little evidence suggested that their breast cancers were detected at earlier stages. The extent of disease at diagnosis did not differ; in situ or localized cancer was diagnosed in 69.0% of women who had no family history of breast cancer and had never received HRT, 73.2% of women who had no family history and had received HRT, 69.4% of women who had a family history and had never received HRT, and 75.0% of women who had a family history and had received HRT (P > 0.2). Some evidence, however, suggested that the tumors detected in women who had ever received HRT were smaller than those detected in women who had never received HRT: Tumors were less than 2 cm in diameter in 71% of the former group and 60% of the latter group (P = 0.002). This finding was slightly more common in women who did not have a family history of breast cancer (72% compared with 67% of women who had a family history; P > 0.2).

Use of Hormone Replacement Therapy, Family History, and Total Mortality

Despite the evidence in the literature that HRT may be associated with a modestly increased risk for breast cancer [14], total mortality rates are decreased with HRT use [8, 14, 15]. Between 1986 and 1993, a total of 2035 deaths were documented among cohort members who did not have cancer at baseline. Among women with no family history of breast cancer, age-adjusted mortality rates ranged from 70 per 10 000 person-years for women who had never received HRT to 51 per 10 000 person-years for women who were currently receiving HRT and had been receiving it for at least 5 years (Table 3). This apparent protective effect of HRT was also evident among women with a family history of breast cancer.

We constructed a multivariate model that included adjustment for age, waist-to-hip ratio, body mass index, alcohol intake, level of physical activity, pack-years of smoking, marital status, hypertension, diabetes, and type of menopause. The multivariate-adjusted relative risks for death were 0.84 (CI, 0.67 to 1.06) in women without a family history of breast cancer and 0.55 (CI, 0.28 to 1.07) in women with a family history. Additional analyses were performed after we excluded participants who had a history of heart disease at baseline and those who died during the first 2 years of follow-up; the point estimates of relative risk were almost identical. Two aspects of the data are worth noting. First, among the women with no family history of breast cancer, the mortality rate in those who currently received HRT and had done so for less than 5 years was 8 deaths per 10 000 person-years per year lower than the rate in women who had never received HRT (62 per 10 000 person-years per year compared with 70 per 10 000 person-years per year). However, adjustment for other risk factors eliminated the apparent benefit (RR, 1.0 [CI, 0.75 to 1.35]). Second, the age-adjusted mortality rates in women who formerly received HRT for at least 5 years (82 per 10 000 person-years per year) was 2 per 10 000 person-years per year higher than the rate in women who had never received HRT (80 per 10 000 person-years per year). However, adjustment for other risk factors produced a relative risk of 0.59 (CI, 0.30 to 1.16).

Use of Hormone Replacement Therapy, Family History, and Cause-Specific Mortality

Mortality rates for specific causes of death were examined to identify whether there were differences by family history (Table 4). Because of the small number of participants in some categories, categories of HRT use were collapsed into "never" and "ever." Use of HRT among women with a family history of breast cancer was associated with decreased rates of death from coronary heart disease, stroke, all cancers combined, and all other causes of death combined. Although it was based on only 84 deaths and was not statistically significant, evidence suggested that women with a family history of breast cancer who used HRT had an increased rate of death from breast cancer (RR, 1.9 [CI, 0.6 to 5.7]). This finding was not entirely unexpected given that mortality rates reflect both incidence and survival. Because it was already known that these women had higher rates of incident breast cancer, we calculated case-fatality rates by using person-years of follow-up from diagnosis until death or end of the follow-up period as the denominator. Compared with the case-fatality rate from breast cancer among women who did not have a family history of breast cancer and had not received HRT, women who had a family history and had received HRT had a 1.3-fold higher (CI, 0.6 to 3.0) risk for death from their disease. This finding, however, was not statistically significant.

Discussion

Top Methods Results Discussion Author & Article Info References

Results from this large prospective cohort study of postmenopausal women suggest that, for most women, HRT is not associated with a statistically significantly increased risk for breast cancer. An exception may be women with a family history of the disease, although the modest increased risk we observed is consistent with chance. What is clear, however, is that women with a family history of the disease who have used HRT seem to have a decreased risk for death. As expected, an inverse association of HRT with mortality rates was evident for coronary heart disease. Evidence suggested an inverse association with the rate of death from stroke and from cancers other than breast cancer. Women with a family history of breast cancer who received HRT were at increased risk for death from breast cancer, but the case-fatality rate was only modestly elevated.

Several studies have examined whether HRT increases the risk for breast cancer [28]. Some of these studies examined the association within strata of other risk factors, including family history. Hoover and colleagues [9] reported that use of conjugated estrogens for 5 years or more was associated with a 3.9-fold increased risk among women with a family history of breast cancer; the corresponding risk estimate among women without a family history was only 1.6. Nomura [29], Wingo [10], and Kaufman [30] and their colleagues reported estimates of effect that were slightly lower but were in the same direction. Brinton and colleagues [31] did not observe a greater risk among the subset with a family history of breast cancer; however, the participants were part of a screening program and the proportion with a family history was high (26%). Several other studies that did not show a greater association among women with a family history reported fewer cases of cancer and may have lacked statistical power to detect an interaction. For example, Mills and colleagues [13] reported a total of only 215 cases, and Palmer and colleagues [32] found only 60 cases of cancer among women with a family history of breast cancer. Newcomb and associates [12] observed a stronger HRT association among women with a family history than we observed, but their results were not statistically significant. The participants in the studies of Yang and colleagues [7] and Stanford and coworkers [5] were younger than our participants. Early-onset breast cancer is more strongly associated with family history than is late-onset breast cancer; the earlier breast cancer develops in relatives, the greater a person's risk [33]. This might explain why our findings differ from those of Yang [7], Stanford [6], and Colditz [11] and their colleagues. The most recent report of the Nurses' Health Study by Colditz and associates [34] did not note differences in risk according to family history of breast cancer.

The availability of data on mammography allowed us to determine whether detection bias may partly explain the increased risk among women who had a family history of breast cancer and had used HRT. Despite the increased use of mammography and apparent increased medical care for these women, the extent of disease at diagnosis did not differ from that in the other subgroups. Some evidence in the literature suggests that women who take estrogens before their diagnosis present with smaller tumors [35, 36]. Our results are consistent with these findings. The tumors found were smaller among women receiving HRT, but this finding was slightly more common among women without a family history of breast cancer (rather than among those with a family history). This trend indicates that detection bias may not be a major explanation for our observation of reduced mortality rates among women who had a family history and received HRT. We excluded women who had cancer at baseline and, for analyses of mortality, excluded women with self-reported heart disease and women who died during the first 2 years of follow-up. Results were unchanged, arguing against the possibility that the lower mortality rate we observed was simply attributable to the fact that healthier women are more likely than unhealthy women to have HRT prescribed [37].

Our study has some limitations that should be considered when our results are being interpreted. First, the family histories of cancer were not validated. However, published validation studies suggest a high degree of accuracy when the site of cancer is the breast and the relatives are first degree [38]. Second, because no data on HRT formulation were available, we could not determine whether risks differ for combined estrogen-progestin HRT. Colditz and colleagues [11] and Newcomb and coworkers [12] reported that the association of HRT with breast cancer does not vary by hormone formulation. Conversely, Stanford and colleagues [6] found that long-term use of combined estrogen-progestin HRT was associated with decreased risk. However, when the Iowa Women's Health Study cohort was established in 1985, combination HRT was probably used by less than 20% of women receiving HRT [39]. Third, we have no data on participants' reasons for taking HRT. In particular, given the considerable HRT-associated reduction in rates of death from coronary heart disease, it would be important to have more detailed information on risk factors for coronary heart disease. Finally, the exposure data on HRT use were collected at baseline and do not necessarily reflect exposure status at time of death or the occurrence of major disease. Because data suggest that the benefit of HRT occurs primarily for women who currently receive and have recently received therapy [40], we may have underestimated the benefit of current HRT use.

Our data have clinical implications for prescription of HRT in women with a family history of breast cancer. As noted earlier, only a handful of published studies of women in this age group have had sufficient statistical power to detect an increased incidence of breast cancer among women who have a family history and use HRT. Grodstein and colleagues [17] examined HRT use and total mortality rates among women with a family history of breast cancer in the Nurses' Health Study and noted a decrease in total mortality rates similar to our findings. Collectively, these results may help physicians and patients with a family history of breast cancer make the undoubtedly difficult decision about whether to begin HRT. Perceptions of one's personal risk might be different for breast cancer and total mortality; however, our data suggest that although risk for breast cancer may be increased with HRT use, risk for death overall is decreased.

Dr. Cerhan: Room 2800 Steinler Building, Preventive Medicine, University of Iowa, Iowa City, IA 52242.