Hydra.usc.edu

Tamoxifen Therapy for Breast Cancer and Endometrial
Cancer Risk
Leslie Bernstein, Dennis Deapen, James R. Cerhan, Stephen M. Schwartz,Jonathan Liff, Erin McGann-Maloney, Jeffrey A. Perlman, Leslie Ford cancer risk was at least as great as that of a 60-year-old woman Background: Tamoxifen is effective in treating breast cancer,
to receive either tamoxifen (20 mg/day for 5 years) or placebo reduces breast cancer incidence among high-risk women,
and accrued more than 13 000 women (5). This study was un- and is associated with increased endometrial cancer risk.
blinded in April 1998 because of the substantial 49% reduced This study was designed to examine the possible modifying
risk of invasive breast cancer among women receiving tamoxi- effects of endometrial cancer risk factors on the tamoxifen-
fen relative to those on the placebo arm. The two European endometrial cancer association. Methods: We conducted a
prevention trials had different eligibility requirements, including case-control study of endometrial cancer (324 case patients
women with different breast cancer risk profiles and permitting and 671 individually matched control subjects) nested within
women to take hormone replacement therapy (HRT) while on a population-based cohort of patients with breast cancer di-
trial. Neither of these trials has shown a benefit for women agnosed from 1978 through 1992 within four regions of the
United States. We obtained information on breast cancer
Tamoxifen has estrogen-like effects in the uterus (8). Reports treatment and endometrial cancer risk factors through in-
of endometrial cancers diagnosed among women receiving terviews and reviews of medical records. All P values re-
tamoxifen therapy for breast cancer began to appear in the lit- ported are two-sided. Results: Endometrial cancer risk was
erature as early as 1985 (9). Elevated endometrial cancer risk has associated with tamoxifen therapy for breast cancer (odds
been confirmed in clinical trials of tamoxifen used for adjuvant ratio = 1.52; 95% confidence interval [CI] = 1.07-2.17). Risk
therapy (10) and population-based studies of breast cancer pa- increased with duration of tamoxifen use (P for trend =
tients (11-14), although the numbers of patients diagnosed with .0002). Women with more than 5 years of exposure to
endometrial cancer in most of the studies are relatively small.
tamoxifen had 4.06-fold greater odds of developing endome-
After reviewing the available animal and human evidence on the trial cancer than nonusers (95% CI = 1.74-9.47). Prior use of
relationship of tamoxifen to the development of endometrial estrogen replacement therapy (ERT) increased risk associ-
cancer, the International Agency for Research on Cancer (15)has classified tamoxifen as a human carcinogen. In the BCPT, ated with tamoxifen use (P for homogeneity of trends
women on the tamoxifen arm had a 2.5-fold greater incidence of <.0001). Risk associated with tamoxifen use was stronger
endometrial cancer than women on the placebo arm (36 invasive among heavier women than among thinner women, although
cancers among women receiving tamoxifen versus 15 invasive trends did not differ statistically (P =.10). Tamoxifen dose-
cancers among women receiving the placebo) (5). All of the response effects were more pronounced among women with
endometrial cancers diagnosed among women on the tamoxifen both previous ERT exposure and higher body mass index
arm of this trial were stage I, and the majority (75%) were than among women in other risk groups. Conclusions: ERT
diagnosed among women who were 50 years of age or older.
use and obesity, both established endometrial cancer risk
Several factors are known to affect endometrial cancer risk, factors and markers of estrogen exposure, substantially
including reproductive characteristics, obesity, use of steroid modify the association between tamoxifen use and endome-
hormone preparations, certain medical conditions, and smoking trial cancer risk among patients with breast cancer. Women
(16). Exposure to estrogen unopposed by progesterone, whether with positive ERT histories and those who are obese, when
endogenous or exogenous, substantially increases women's risk prescribed tamoxifen, may warrant closer surveillance for
of this disease. Use of combination oral contraceptive prepara- endometrial cancer than women without such histories. [J
tions substantially lowers risk. Obesity, a source of endogenous Natl Cancer Inst 1999;91:1654-62]
Tamoxifen, a nonsteroidal hormone that acts as an antiestro- Affiliations of authors: L. Bernstein, D. Deapen, E. McGann-Maloney, De- partment of Preventive Medicine, University of Southern California School of gen in breast tissue, was approved by the U.S. Food and Drug Medicine, and Norris Comprehensive Cancer Center, Los Angeles; J. R. Cerhan, Administration for the treatment of advanced breast cancer Department of Health Sciences Research, Mayo Clinic, Rochester, MN; S. M.
among postmenopausal women in 1978. Currently, tamoxifen is Schwartz, Division of Public Health Sciences, Fred Hutchinson Cancer Research used among women of all ages for the treatment of all stages of Center, and Department of Epidemiology, School of Public Health and Com- breast cancer (1). Tamoxifen reduces the risk of subsequent munity Medicine, University of Washington, Seattle; J. Liff, Department of contralateral breast cancer as well as breast cancer recurrences Epidemiology, Rollins School of Public Health, Emory University, Atlanta, GA;J. A. Perlman, U.S. Public Health Service and Epimedix, Washington, DC; L.
and mortality (2-4). Because of its efficacy in breast cancer Ford, Division of Cancer Prevention, National Cancer Institute, Bethesda, MD.
therapy, clinical trials were initiated in the United States, the Correspondence to: Leslie Bernstein, Ph.D., Department of Preventive Medi- U.K., and Italy among disease-free women to evaluate the effi- cine, University of Southern California/Norris Comprehensive Cancer Center, cacy of tamoxifen in the primary prevention of breast cancer.
1441 Eastlake Ave., MS 44, Los Angeles, CA 90033 (e-mail: lbern@hsc.
The National Surgical Adjuvant Breast and Bowel Project (NSABP) Breast Cancer Prevention Trial (BCPT), which is the See "Notes" following "References." largest of these trials, randomly assigned women whose breast- Journal of the National Cancer Institute, Vol. 91, No. 19, October 6, 1999 (unopposed) estrogen among postmenopausal women, increases identified 330 eligible case patients and 708 eligible control subjects. A total of endometrial cancer risk. None of the prior studies of tamoxifen 397 control subjects (56%) matched case patients exactly on all matching cri- therapy and endometrial cancer risk has adequately considered whether these risk factors modify the tamoxifen-endometrial Collection of Treatment Histories and Risk Factor
Information
We designed a population-based, case-control study, nested within the cohort of breast cancer patients diagnosed within four Detailed information on all treatments (surgery, radiation therapy, chemo- geographically defined regions served by four Surveillance, Epi- therapy, and hormonal therapy) that a woman received for her breast cancer was demiology, and End Results (SEER)1 registries, to examine the abstracted from hospital medical records and the records of all physicians pro-viding such treatment. Special efforts were made to ensure that any treatment relationship between tamoxifen use (including duration of use, given for progression of disease, recurrence of disease, or second primary breast recentness of use, and cumulative dose) and subsequent devel- cancer diagnosis was abstracted. In addition, information on age at menopause, opment of endometrial cancer. This study was also designed to parity, family history of breast cancer, and other medical conditions, such as address whether estrogen replacement therapy (ERT), oral con- diabetes, hypertension, stroke, and coronary heart disease, was obtained from the traceptive use, and obesity modify any observed relationship.
patient's medical records. Height and weight were obtained from medical re-cords at the woman's admission physical at the time of her initial breast cancer SUBJECTS AND METHODS
Patients alive at the time of the study were interviewed by telephone to obtain Subject Identification and Eligibility
further information on relevant endometrial cancer risk factors and breast cancertherapy. Women were asked about their reproductive histories, medical history, All women diagnosed with breast cancer who had had no prior or concurrent smoking history, and use of oral contraceptives, ERT, or combined HRT (regi- cancers (other than bilateral breast cancer) at the time of initial breast cancer mens of estrogen and a progestin). All women were asked to provide a roster of diagnosis were identified at the four SEER registries. Patients first diagnosed physicians who had provided their health care during their adult years, including from 1978 through 1992 were eligible for the study in Los Angeles County, general or family practitioners, internists, cardiologists, gynecologists, oncolo- whereas for the other SEER registries, the years of initial breast cancer diagnosis gists, and surgeons. In Los Angeles County, next of kin were also interviewed if were 1983 through 1988 for Atlanta (GA), 1983 through 1990 for Iowa, and the patient was deceased or unable to respond to the interview. We conducted 1983 through 1989 for Seattle-Puget Sound (WA). Each of these SEER regis- 227 next-of-kin interviews (case patients-86 or 37% of the 232 eligible pa- tries is a population-based cancer registry serving a designated geographic region tients; control subjects-141 or 27% of the 521 eligible patients). At all sites, as part of the National Cancer Institute's Cancer Statistics Program. Case pa- medical records were sought from all physicians mentioned in the interview, tients were women diagnosed with endometrial cancer at least 6 months after an whether done with patients or next of kin as well as any physician mentioned in initial breast cancer diagnosis within the defined period. Case patients had no the hospital or other physician record. At sites other than Los Angeles County, prior cancer diagnoses and no cancer diagnosed between their breast cancer when we were unable to conduct a patient interview, we relied completely on diagnosis and their endometrial cancer diagnosis other than a second primary patient medical records for data collection.
breast cancer (or basal or squamous cell skin cancer). The years of endometrial At the beginning of the telephone interview, each participating subject pro- cancer diagnosis for subjects considered to be eligible for this study were 1978 vided informed consent. Study procedures were approved by institutional review through June 1993 in Los Angeles County and 1983 through 1991 in Atlanta, boards at the University of Southern California (Los Angeles), University of Iowa, and Seattle-Puget Sound. Within the latter three registries, no eligible case Iowa (Iowa City), Fred Hutchinson Cancer Research Center (Seattle, WA), and patient had an endometrial cancer diagnosis in 1983.
Emory University (Atlanta, GA), in accord with assurances approved by the U.S.
Control subjects were breast cancer patients who did not develop endometrial Department of Health and Human Services.
cancer and who were selected individually for each case patient to be comparable On the basis of all sources of information, we reconstructed a detailed medical to the case patient with respect to specific characteristics. We individually history for each patient. In compiling all sources of information, we sought matched two control subjects to each case patient on the year of first breast medical record confirmation for all drug exposures and breast cancer treatments.
cancer diagnosis, year of birth, race/ethnicity (non-Hispanic white, Hispanic Although we were able to confirm most exposure histories of women with white, black, or Asian), SEER registry, and summary stage of disease (localized, next-of-kin interviews through review of medical records, we were unable to regional extension, or metastatic disease). The duration of time that a case patient confirm positive histories of tamoxifen use for eight of these patients, negative was at risk for endometrial cancer was calculated as the number of months histories of tamoxifen use for 39 of these patients, positive histories of hormone between her initial breast cancer diagnosis and her endometrial cancer diagnosis; use for 18 of these patients, and negative histories of such use for 61 of these each control subject was required to have survived at least the same length of patients. The compiled medical history included dosages of all chemotherapy time as her matched case patient without any subsequent cancer diagnosis other regimens and the dates that they were administered; dates, duration, and dosages than a second primary breast cancer (or basal or squamous cell skin cancer) and of tamoxifen therapy; details of radiation therapy; and dates and duration of ERT to have had an intact uterus on the last day of that follow-up period. Control and HRT. For oral contraceptive use, we were able to determine whether women subjects were required to have maintained their residence within the geographic had used this method of contraception but were unable to obtain details on the area covered by the registry so that, had they become case patients, the registry duration of use for most women. We collected information on breast cancer would have ascertained their subsequent cancer.
therapies as well as the use of exogenous hormones throughout the defined From a roster of all patients with breast cancer eligible as potential matches for follow-up period based on the number of days between the case patients' breast each case patient, we randomly selected control subjects and confirmed their cancer and endometrial cancer diagnoses. We included use of estrogens by pill, eligibility by use of hospital and physician medical records and interview infor- patch, or injection in the category of ERT; some of this use occurred during the mation so that at least two control subjects satisfying all eligibility criteria were women's premenopausal and perimenopausal years. To determine whether a selected. Where we lacked sufficient control subjects who were exact matches woman had used ERT, we required that there be no more than a 2-month lapse for a particular case, we first relaxed the matching criteria of year of birth to year in continuous medical records. Otherwise, we considered that the history of ERT of birth within 1 or 2 years; if necessary, we also relaxed the year of diagnosis use was unknown. It is likely that most women so designated had never used matching criteria to year of diagnosis within 1 year. The hysterectomy status and residential history of each potential control subject were established throughmedical record reviews and telephone interviews.
Statistical Analyses
In Los Angeles County, we identified more than two control subjects for some case patients as, initially, we randomly selected five to 10 potential control The 330 eligible case patients included 232 from Los Angeles County, 50 subjects for each case patient and initiated data collection for the first five. We from Iowa, 34 from Seattle-Puget Sound, and 14 from Atlanta. The majority of expected frequent losses because of ineligibility by virtue of a control subject these women were non-Hispanic whites (n ס 305); 14 were Hispanic whites, six having had a prior hysterectomy. At other study sites, we only attempted to were African-Americans, and five were Asian-Americans. A total of 708 control identify two control subjects per case patient. Across the four study sites, we subjects were determined to be eligible matches for these 330 case patients. We Journal of the National Cancer Institute, Vol. 91, No. 19, October 6, 1999 retained all eligible control subjects in the statistical analyses. Over all study diagnosis, and history of high blood pressure at breast cancer diagnosis except sites, we were unable to identify a suitable control subject for one case patient where a different form of one of these variables was being evaluated. Tests for and could not determine whether five case patients had taken tamoxifen. These trend were computed by fitting a conditional logistic regression model to con- six case patients and their 10 individually matched control subjects were ex- tinuous values of the variables. All reported trend test significance levels (P cluded from all analyses. In addition, we excluded 27 eligible control subjects values) are two-sided. To assess whether other endometrial cancer risk factors who had missing information on tamoxifen exposure. Exclusion of these control statistically significantly modified the effects of tamoxifen on endometrial can- subjects did not result in the exclusion of any case patients because each had at cer risk, we constructed a likelihood ratio test to determine homogeneity of least one remaining eligible matched control subject. Thus, the statistical analy- trends in risk with increasing levels of tamoxifen exposure.
ses are based on 324 case patients (98% of total eligible patients) and 671 control Analyses of the effects of tamoxifen on endometrial cancer risk that were subjects (96% of total eligible matches for the 324 case patients) with a matching conducted within strata of exposure variables modeled all levels of the stratifi- ratio ranging from one to four. The distribution of patient characteristics is cation variable simultaneously and excluded women who were missing infor- mation on the stratification variable or the duration of tamoxifen use unless Quetelet's index (weight in kilograms divided by height in meters squared) otherwise indicated. For these analyses, we used unconditional logistic regres- was used as a measure of body mass index. Exposure to tamoxifen was ex- sion analyses and adjusted for all factors on which we matched in the study pressed as the total duration of exposure in months and as cumulative dose in design. We also conducted analyses restricted to women who were postmeno- milligrams. Exposure to ERT or HRT was expressed as the total duration of pausal at the time of their breast cancer diagnoses by use of the same statistical exposure in months. When creating analytic variables for ERT and HRT, we restricted the referent group to women who had not used either type of regimen.
We classified women according to their smoking status at the time of breastcancer diagnosis as current smokers or current nonsmokers.
The majority of the case patients were diagnosed with local- Univariate and multivariate conditional logistic regression methods with a ized (61.1%) or regional (37.3%) breast cancer. Eighteen case variable number of control subjects matched to each case patient were used to patients (5.6%) and 13 control subjects (1.9%) were diagnosed estimate odds ratios (ORs) and 95% confidence intervals (CIs) for the OR (17). with a second primary breast cancer prior to their endometrial In all categorical variable analyses, women with missing information for a par-ticular variable were included in a separate category in the analysis. All multi- cancer diagnosis (case patients) or the end of their at-risk period variate analyses included duration of tamoxifen therapy, duration of ERT, any (control subjects). A total of 70 case patients (21.6%) and 149 use of oral contraceptives, body mass index, smoking status at breast cancer control subjects (22.2%) had recurrent or metastatic disease. Theaverage age at breast cancer diagnosis was 65.9 years (range, Table 1. Characteristics of study population
38.4-92.3 years) for case patients and 65.6 years (range, 38.8-93.6 years) for control subjects. The time interval between breast cancer diagnosis and endometrial cancer diagnosis for case pa- tients averaged 3.9 years (range, 6 months to 13.5 years). One hundred case patients (and their 205 matched control subjects) had at least a 5-year at-risk interval (Table 1).
Women who used oral contraceptives were at modestly re- duced risk of endometrial cancer relative to nonusers (OR ס 0.59; 95% CI ס 0.35-1.01) (Table 2). Any use of ERT was associated with a twofold increased risk (OR ס 2.12; 95% CI ס 1.52-2.96). Women with more than 8 years of ERT use had more than four times the risk of unexposed women. Of note, 14 case patients and nine control subjects used ERT following theirbreast cancer diagnoses. We determined how recently each ERT-exposed patient had used ERT. We observed an elevated risk for any ERT use within 5 years of endometrial cancer di- agnosis or the end of the follow-up period (for control subjects) as well as for ERT use that ended more than 5 years before the endometrial cancer diagnosis or the end of the follow-up period (Table 2). The OR estimates for the two exposure groups did not Few women (23 case patients and 47 control subjects) were known to have used HRT; of these, 17 case patients (74%) and 23 control subjects (49%) had previously used ERT. We also documented that three case patients and nine control subjects used HRT following their breast cancer diagnoses. Overall, HRT use was associated with a modest elevation in endometrial can- cer risk (OR ס 1.69; 95% CI ס 0.93-3.06). However, among those women with no prior ERT exposure, HRT use was not associated with increased endometrial cancer risk (OR 95% CI ס 0.29-2.06) in a multivariate model.
Approximate quartile categories were created for body mass index on the basis of the distribution of this index among control subjects. Endometrial cancer risk increased with increasing cat- egory of body mass index. Women in the highest category had a twofold greater risk than women in the lowest category (OR ס *SEER ס Surveillance, Epidemiology, and End Results.
2.06; 95% CI ס 1.31-3.24) (Table 2).
Journal of the National Cancer Institute, Vol. 91, No. 19, October 6, 1999 Table 2. Odds ratios (ORs) and 95% confidence intervals (95% CIs) of endometrial cancer associated with selected exposures among
patients previously diagnosed with breast cancer History of high blood pressure at breast cancer diagnosis History of diabetes at breast cancer diagnosis Smoking status at breast cancer diagnosis *All multivariate models include categorical terms for months of tamoxifen therapy, months of estrogen replacement therapy, use of hormone replacement therapy only, oral contraceptive use, body mass index (body weight in kilograms divided by height in meters squared), smoking status at diagnosis, and history of high bloodpressure at diagnosis, except where a different formulation of one of these variables was included in the model.
†Tests for trend were computed by fitting conditional logistic regression models to continuous values of the variables.
‡Referent group ס women with no use of estrogen replacement therapy or combined hormone replacement therapy.
We restricted the history of other medical conditions to those Neither chemotherapy nor radiation therapy for breast cancer diagnosed prior to the patient's diagnosis of breast cancer. En- was associated with an elevated risk of endometrial cancer dometrial cancer risk was statistically significantly elevated (Table 3). Few women in this study received radiation therapy to among women with a history of high blood pressure (OR ס 1.58; 95% CI ס 1.20-2.08) and nonsignificantly elevated Women treated with tamoxifen had a greater risk of endo- among those with a history of diabetes (OR ס 1.36; 95% CI ס metrial cancer than those who did not take tamoxifen (OR ס 0.91-2.05) (Table 2). Adjustment for body mass index reduced 1.52; 95% CI ס 1.07-2.17) after multivariate adjustment (Table the OR associated with diabetes to 1.23 (95% CI ס 0.81-1.86), 3). Risk increased 18% per year of use and was statistically although it increased to 1.33 following adjustment for other significantly elevated among women who were treated with factors in the multivariate model. After adjustment for body tamoxifen for more than 2 years. We also show results for the mass index, the OR for high blood pressure, although attenuated total cumulative dose of tamoxifen a woman received. The ma- somewhat, remained statistically significant (OR ס 1.43; 95% jority of women received 20 mg/day throughout their treatment, CI ס 1.08-1.91). The OR for high blood pressure was reduced although some had their doses altered during the course of their minimally after adjustment for other potential confounding fac- treatment and a few received other doses (10, 30, or 40 mg/day).
tors. Endometrial cancer risk was not associated with a history of Among women taking tamoxifen, cumulative dose and duration coronary heart disease or stroke (data not shown). Women who of use were highly correlated (Pearson r ס.99 for case patients were current cigarette smokers had a nonsignficant, reduced risk and r ס.97 for control subjects). Therefore, the risk estimates of endometrial cancer relative to current nonsmokers (OR ס for cumulative dose are similar to those for duration of therapy 0.74; 95% CI ס 0.52-1.05) (Table 2).
Journal of the National Cancer Institute, Vol. 91, No. 19, October 6, 1999 Table 3. Odds ratios (ORs) and 95% confidence intervals (95% CIs) of endometrial cancer associated with treatment of breast cancer
Recentness and duration of tamoxifen therapy *All multivariate models include categorical terms for months of tamoxifen therapy, months of estrogen replacement therapy, use of hormone replacement therapy only, oral contraceptive use, body mass index (body weight in kilograms divided by height in meters squared), smoking status at diagnosis, and history of high bloodpressure at diagnosis, except where a different formulation of one of these variables was included in the model.
†Tests for trend were computed by fitting conditional logistic regression models to continuous values of the variables.
Most women with tamoxifen exposure were currently taking dometrial cancer (OR ס 1.62; 95% CI ס 1.05-2.50), whereas tamoxifen or had taken it within 12 months of the end of the women who had used both drugs were at substantially increased at-risk period. We combined use within 12 months with current risk of endometrial cancer (OR ס 3.53; 95% CI ס 2.15-5.78).
use into one category because case patients may have experi- We further examined the patterns of use among women exposed enced symptoms such as bleeding prior to their endometrial to both tamoxifen and ERT. Among those women who were cancer diagnoses, which might have resulted in their being ad- treated with tamoxifen within 1 year of last use of ERT (19 case vised to discontinue tamoxifen. Although having last used patients and 12 control subjects), the relative odds of endome- tamoxifen more than 12 months ago was not associated with trial cancer was 5.28 (95% CI ס 2.35-11.9); the risk estimate endometrial cancer risk, this may be masking a duration effect was lower for women with more than 1 year between the use of since, among women with more than 2 years of tamoxifen use, both regimens (30 case patients, 31 control subjects; OR ס the ORs for recent and past users were of similar magnitude 2.83; 95% CI ס 1.51-5.31). On the basis of a test for homogen- (twofold increase in risk) (Table 3).
eity, these risk estimates do not differ statistically (P ס.18).
We examined the combined effects of tamoxifen therapy and The trend in risk associated with increasing duration of ERT on the risk of endometrial cancer. Comparing women on tamoxifen therapy was not statistically significant among the basis of whether they had ever used either drug, we observed women who had never used ERT (P ס.17) (Table 4); women no increased risk of endometrial cancer among those women who used tamoxifen for more than 60 months had an elevated who had ever used tamoxifen but had never used ERT relative to risk, but the CI for the OR includes 1.0. In contrast to these women unexposed to either drug (OR ס 1.14; 95% CI ס 0.73- results, among women who had previously used ERT, tamoxifen 1.81). Those using only ERT had a 60% increased risk of en- therapy was strongly and statistically significantly associated Journal of the National Cancer Institute, Vol. 91, No. 19, October 6, 1999 Table 4. Multivariate odds ratios (ORs) and 95% confidence intervals (95% CIs) of endometrial cancer associated with tamoxifen
treatment of breast cancer in subgroups of women defined by other endometrial cancer risk factors* Any use of estrogen replacement therapy (ERT) No. of case patients/No. of control subjects No. of case patients/No. of control subjects No. of case patients/No. of control subjects No. of case patients/No. of control subjects *Excluding the exposure(s) of interest, the multivariate models include categorical terms for months of estrogen replacement therapy, use of hormone replacement therapy only, oral contraceptive use, body mass index (body weight in kilograms divided by height in meters squared), smoking status at diagnosis, and history ofhigh blood pressure at diagnosis.
†Tests for trend were computed by fitting unconditional logistic regression models to continuous values of the variables.
with endometrial cancer risk (P for trend <.0001). Relative to women: 1) those who had never used ERT and were thinner (i.e., women not treated with tamoxifen, those who used tamoxifen below the median body mass index of control subjects), 2) those for 1-12 months had 1.8 times the risk of endometrial cancer.
who had never used ERT and were heavier (i.e., above the The ORs increased substantially with increasing duration of use median body mass index of control subjects), 3) those who had and, among women who received tamoxifen for more than 60 used ERT but were thinner, and 4) those who had used ERT and months, the relative odds of endometrial cancer was 5.73 (95% were heavier. Tamoxifen use was associated with a much greater CI ס 1.64-20.0). On the basis of the test for homogeneity of increase in risk of endometrial cancer among heavier women trends, the dose-response effects of tamoxifen on endometrial who had previously used ERT than among women in the other cancer risk were statistically significantly different for users and categories of ERT use and obesity (test for homogeneity of never users of ERT (P<.0001). Among women with missing trends, P ס.0008) (Table 4). Among women with no ERT ERT information, the OR was elevated only among those with exposure, the OR estimates are quite similar for those with low more than 60 months of tamoxifen use.
and with high body mass. Fitting a single trend for these two The trends in risk of endometrial cancer associated with subgroups provides an equivalent fit of the data with one fewer tamoxifen therapy did not differ statistically between women with low versus high body mass index (test for homogeneity of The trends in risk by smoking status did not differ statistically trends, P ס.10) (Table 4). For heavier women, those who had (test for homogeneity of trends, P ס.23) (Table 4).
used tamoxifen for more than 60 months had nearly a fivefold Women who had experienced their last menstrual period at greater risk of endometrial cancer than those who had not been least 1 year prior to their breast cancer diagnosis were consid- treated with tamoxifen (OR ס 4.98; 95% CI ס 1.95-12.7).
ered to be postmenopausal. We were unable to determine di- Among thinner women, this risk was 2.4-fold greater (OR ס rectly the menopausal status of 112 women in the study; we classified 107 women with unknown status who were 56 years We examined the risk of endometrial cancer in four groups of old or older at the time of their breast cancer diagnosis as post- Journal of the National Cancer Institute, Vol. 91, No. 19, October 6, 1999 menopausal. Ninety-four women were known to be premeno- ERT exposure, only those with more than 5 years of tamoxifen pausal at the time of their breast cancer diagnosis; the remaining exposure appear to be at increased risk of endometrial cancer, five with unknown menopausal status were under age 50 years although we cannot rule out a similar twofold increased risk with when diagnosed with breast cancer and were considered to be use of 2-5 years. Similarly, the trend in risk with increasing premenopausal. The 99 premenopausal women included 32 case duration of tamoxifen therapy was greater among women with patients and 67 control subjects. Among these 99 women, nine higher body mass index (>24.5 kg/m2) than among women with had used ERT (four case patients and five control subjects), six lower body mass index, although we cannot rule out increased had used HRT (one case patient and five control subjects), and risk with long duration of use among thinner women. Addition- 29 had received tamoxifen (seven case patients and 22 control ally, the combined modifying effects of prior ERT exposure and subjects). In an analysis restricted to postmenopausal women, body mass index suggest that women with prior ERT exposure risk estimates for exposure to ERT, HRT, and tamoxifen therapy and high body mass index are at the greatest risk of endometrial did not differ substantially from those observed for all study cancer in association with tamoxifen treatment of breast cancer.
subjects except for subjects in the longest duration of ERT- and This analysis of all three factors (ERT exposure, body mass tamoxifen-use categories. The OR for women who had used index, and tamoxifen exposure) also suggests that, among ERT for more than 96 months was lowered by 8% and that for women with low body mass index, ERT and tamoxifen combine women who had used tamoxifen for more than 60 months was to increase endometrial cancer risk, whereas among women with no ERT exposure, the effects of tamoxifen are similar in both ISCUSSION
ERT is an acknowledged risk factor for endometrial cancer Tamoxifen therapy for breast cancer was associated with a on the basis of results from cohort and case-control studies that fourfold increased risk of endometrial cancer among women show a strong, persistent duration response relationship (with with more than 5 years of exposure, a duration of treatment that risk estimates of 1.4 for up to 1 year of use, 2.8 for 1-5 years of is no longer recommended (18). Tamoxifen was associated with use, 5.9 for 5-10 years of use, and 9.5 for more than 10 years of a more modest increase in risk for women with 2-5 years of use use that remain elevated more than 5 years after cessation of use) and was unrelated to risk among women with a shorter duration [reviewed in (16)]. Our risk estimates are somewhat lower than these; this may reflect the fact that a majority of our exposed That tamoxifen increases the risk of endometrial cancer is women had not used ERT for many years prior to their endo- consistent with data on the known estrogen-like effects of metrial cancer diagnoses (average: 10.2 years since last use for tamoxifen in women. Tamoxifen appears to have selective af- case patients and 12.2 years since last use for control subjects).
finity for endometrial tissue (19) and, among premenopausal However, we did examine whether recent use (within 5 years) or women, acts directly on the ovaries to stimulate estrogen bio- more remote use affected the magnitude of the association with synthesis and increase plasma estrogen levels (20,21). Among ERT; both risk estimates were elevated and statistically consis- postmenopausal women, although tamoxifen may have a small suppressive effect on circulating estrogen levels (22), this low Endometrial cancer risk is also related to obesity, although estrogenic environment may permit the activation and increased studies are inconsistent as to whether a dose-response rela- synthesis of endometrial estrogen and progesterone receptors by tionship exists or whether only the most obese women are at tamoxifen (23,24). Tamoxifen also has been shown to cause increased risk (16). In our study, endometrial cancer risk in- estrogen-like changes in the vaginal epithelium (25) and endo- creased with increasing level of body mass index as measured metrium (26) of some women. Tamoxifen is also associated with by Quetelet's index. The relationship between obesity and en- endometrial thickening, endometrial hyperplasia, and endocer- dometrial cancer risk may also represent the effects of greater vical and endometrial polyps (27,28) and thus may have a direct estrogen exposure on the uterus. Among obese postmenopausal stimulatory effect on the uterine body and endometrium (27). women, the peripheral conversion of androstenedione to estrone Women with pre-existing, asymptomatic endometrial lesions are is a major source of estrogen (31). Since circulating levels of more likely than women with no endometrial abnormalities to sex hormone-binding globulin are inversely related to obesity, develop atypical lesions while on tamoxifen therapy (29). Al- this results in higher levels of unbound (and therefore, bioavail- though estrogen-like effects of tamoxifen on the endometrium able) estrogen during ages when such levels are generally low are the most likely explanation for increased endometrial cancer risk, other pathways, such as the regulation of insulin-like Our result showing about a 40% reduction in endometrial growth factor-I, may also play a role (30). cancer risk associated with oral contraceptive use is consistent Our results suggest that the relationship between tamoxifen with the existing literature on this topic (16). We were unable to use and subsequent endometrial cancer risk is substantially collect detailed data on formulation or duration of use. Cigarette modified by ERT use and body mass index. In the absence of smoking and endometrial cancer risk are inversely related in these exposures, tamoxifen may be associated with only a mod- many studies, possibly through an antiestrogenic mechanism est increase in risk. We find a relationship between tamoxifen (16). We observe a small reduction in risk among women who therapy for breast cancer and endometrial cancer risk that is were current smokers at the time of their breast cancer diag- substantially greater among women with prior exposure to ERT noses. Other studies have reported an increased risk of endome- and among those with high body mass index at breast cancer trial cancer among women with diabetes and hypertension, al- diagnosis than among those without such exposures. ORs in- though the majority did not consider potential confounding creased dramatically among women with ERT exposure, and factors, such as obesity (for both conditions) and ERT (for hy- even short-term tamoxifen exposures (ഛ24 months) were asso- pertension) (16). We, too, observed a modest elevated risk as- ciated with increased risk. Among women without a history of sociated with these histories, with the association with hyper- Journal of the National Cancer Institute, Vol. 91, No. 19, October 6, 1999 tension remaining statistically significantly elevated following potential impact on this relationship of confounding and effect modification by acknowledged endometrial cancer risk factors, Our results evaluating the main effect of tamoxifen on endo- particularly ERT exposure and obesity.
metrial cancer risk are consistent with the majority of case- The issue of possible detection bias noted by MacMahon (33) control studies in the literature showing that increasing duration has been raised with regard to the relationship between ERT and of tamoxifen therapy for breast cancer results in a graded in- endometrial cancer risk (34,35). Since tamoxifen can cause gy- creasing risk of endometrial cancer (12-14). One negative study necologic symptoms, women receiving tamoxifen are often in- (32) had limited duration of tamoxifen exposure among its par- vestigated by transvaginal ultrasonography and hysteroscopy, ticipants. The cumulative total number of cases included in these resulting in the diagnosis of occult cancers. Although the first previous case-control studies was only 311, with 97 cases in the report of such a relationship between tamoxifen therapy and Dutch study (12), 36 cases in the negative U.S. study (32), and endometrial cancer risk was published in 1985 (9), the major 43 cases in the first and 135 cases in the second French study reports did not appear in the literature until 1994 (12) or later.
(13,14). Only the U.S. case-control study (32) attempted to Therefore, detection of occult endometrial cancers as a result of evaluate potential confounding factors. In the European studies increased screening of asymptomatic women is not likely to (12-14), substantially more women were exposed to doses of have affected our study because case patients were diagnosed 30-40 mg/day than we observed in our study where nearly all with endometrial cancer between 1980 and mid-1993.
women were treated with 20 mg/day. Our risk estimates are In summary, this study confirms that tamoxifen therapy for quite similar to those from the Dutch study (12) and from the breast cancer is associated with an increased risk for endometrialcancer. It further demonstrates that the strength of the relation- study by Sasco et al. (13) (the first French study). Although ship is substantially affected by the woman's history of exposure Mignotte et al. (14) (the second French study) had higher risk to unopposed exogenous estrogens and her body mass index at estimates than ours, their CIs are wide and consistent with our the time of breast cancer diagnosis. In the absence of prior ERT risk estimates. Mignotte and colleagues compared their results exposure or obesity, the effects of tamoxifen on risk are consid- with those from the Dutch study (12) and suggested that risk was erably lower than those observed among women with these ex- higher among their patients because their patients had accumu- posures. These results are consistent with tamoxifen having an lated greater durations of exposure and larger cumulative doses estrogenic effect on the endometrium, enhancing the effects of of tamoxifen than the Dutch women. Our results suggest that ERT and obesity. Because women with prior ERT use and those other factors may explain these differences, particularly history who are heavier appear to have a greater risk of endometrial of exposure to ERT and body mass index at the time of breast cancer than women without these exposures, physicians should be particularly vigilant in monitoring tamoxifen-treated patients A study based on breast cancer patients' first course of treat- with these additional risk factors. Because tamoxifen has proven ment and subsequent primary cancer diagnoses collected by the benefits in extending the disease-free and overall survival of SEER registries also showed a statistically significant twofold breast cancer patients and in reducing the incidence of breast elevation in endometrial cancer risk associated with hormonal cancer among women at increased risk, it remains an important therapy (11). These results are not definitive as to treatment therapeutic option for women with all stages of breast cancer as because they do not include any therapy after the first course and well as for healthy women who are at increased risk of breast breast cancer cases classified in the tamoxifen group may have received other hormonal therapies in addition to or other thantamoxifen.
REFERENCES
Breast cancer clinical trials of tamoxifen have shown mixed results with regard to the tamoxifen-endometrial cancer rela- (1) Jordan VC. Tamoxifen: the herald of a new era of preventive therapeutics tionship; some demonstrated an increased risk of endometrial [editorial]. J Natl Cancer Inst 1997;89:747-9.
cancer, whereas others showed no association (33). The NSABP (2) Early Breast Cancer Trialists' Collaborative Group. Systemic treatment of B-14 trial results on endometrial cancer risk showed a relative early breast cancer by hormonal, cytotoxic, or immune therapy. 133 ran- risk of 7.5 when tamoxifen-treated patients were compared with domised trials involving 31,000 recurrences and 24,000 deaths among75,000 women. Lancet 1992;339:1-15.
control subjects, who may have had a deficit of endometrial (3) Early Breast Cancer Trialists' Collaborative Group. Tamoxifen for early cancer, and 2.3 when tamoxifen-treated patients were compared breast cancer: an overview of the randomised trials. Lancet 1998;351: with patients participating in another trial (B-06) (10). Although this trial collected information on HRT use at the time of study (4) Nayfield SG, Karp JE, Ford LG, Dorr FA, Kramer BS. Potential role of entry, these data were not verified nor was information collected tamoxifen in the prevention of breast cancer. J Natl Cancer Inst 1991;83: on type of drug (ERT versus HRT), duration of use, or recent- (5) Fisher B, Costantino JP, Wickerham DL, Redmond CK, Kavanah M, Cro- nin WM, et al. Tamoxifen for prevention of breast cancer: report of the MacMahon (33) reviewed all studies of the relationship be- National Surgical Adjuvant Breast and Bowel Project P-1 Study. J Natl tween tamoxifen use and endometrial cancer risk, covering re- ports published prior to 1997. He concluded that, although these (6) Veronesi U, Maisonneuve P, Costa A, Sacchini V, Maltoni C, Robertson C, studies taken together suggest an association, the studies do not et al. Prevention of breast cancer with tamoxifen: preliminary findings from adequately address potential confounding factors and the clinical the Italian randomised trial among hysterectomised women. Italian trials do not address the issue of detection (unmasking) bias. He Tamoxifen Prevention Study. Lancet 1998;352:93-7.
(7) Powles T, Eeles R, Ashley S, Easton D, Chang J, Dowsett M, et al. Interim suggests that case-control studies can better address these issues analysis of the incidence of breast cancer in the Royal Marsden Hospital than clinical trials. Our study was designed with sufficient tamoxifen randomised chemoprevention trial. Lancet 1998;352:98-101.
sample size and detailed data collection procedures, including (8) Assikis VJ, Jordan VC. Risks and benefits of tamoxifen therapy. Oncology extensive medical record review and interviews, to evaluate the Journal of the National Cancer Institute, Vol. 91, No. 19, October 6, 1999 (9) Killackey MA, Hakes TB, Pierce VK. Endometrial adenocarcinoma in (26) Boccardo F, Guarneri D, Rubagotti A, Casertelli GL, Bentivoglio G, Conte breast cancer patients receiving antiestrogens. Cancer Treat Rep 1985;69: N, et al. Endocrine effects of tamoxifen in postmenopausal breast cancer (10) Fisher B, Costantino JP, Redmond CK, Fisher ER, Wickerham DL, Cronin (27) Kedar RP, Bourne TH, Powles TJ, Collins WP, Ashley SE, Cosgrove DO, WM, et al. Endometrial cancer in tamoxifen-treated breast cancer patients: et al. Effects of tamoxifen on uterus and ovaries of postmenopausal women findings from the National Surgical Adjuvant Breast and Bowel Project in a randomised breast cancer prevention trial. Lancet 1994;343:1318-21.
(NSABP) B-14. J Natl Cancer Inst 1994;86:527-37.
(28) Lahti E, Blanco G, Kauppila A, Apaja-Sarkkinen M, Taskinen PJ, Laati- (11) Curtis RE, Boice JD Jr, Shriner DA, Hankey BF, Fraumeni JF Jr. Second kainen T. Endometrial changes in postmenopausal breast cancer patients cancers after adjuvant tamoxifen therapy for breast cancer. J Natl Cancer receiving tamoxifen. Obstet Gynecol 1993;81(Pt 1):660-4.
(29) Berliere M, Charles A, Galant C, Donnez J. Uterine side effects of tamoxi- (12) van Leeuwen FE, Benraadt J, Coebergh JW, Kiemeney LA, Gimbrere CH, fen: a need for systematic pretreatment screening. Obstet Gynecol 1998; Otter R, et al. Risk of endometrial cancer after tamoxifen treatment of breast cancer. Lancet 1994;343:448-52.
(30) Elkas J, Gray K, Howard L, Petit N, Pohl J, Armstrong A. The effects of (13) Sasco AJ, Chaplain G, Amoros E, Saez S. Endometrial cancer following tamoxifen on endometrial insulin-like growth factor-I expression. Obstet breast cancer: effect of tamoxifen and castration by radiotherapy. Epide- (31) Anderson DC. Sex-hormone-binding globulin. Clin Endocrinol (Oxsf) (14) Mignotte H, Lassett C, Bonadona V, Lesur A, Luporsi E, Rodier JF, et al.
Iatrogenic risks of endometrial carcinoma after treatment for breast cancer (32) Cook LS, Weiss NS, Schwartz SM, White E, McKnight B, Moore DE, et in a large French case-control study. Int J Cancer 1998;78:325-30.
al. Population-based study of tamoxifen therapy and subsequent ovarian, (15) International Agency for Research on Cancer (IARC), editor. Tamoxifen.
endometrial, and breast cancers. J Natl Cancer Inst 1995;87:1359-64.
IARC Monographs on the Evaluation of Carcinogenic Risks to Humans.
(33) MacMahon B. Overview of studies on endometrial cancer and other types Vol. 66. Lyon (France): IARC; 1996. p. 253-365.
of cancer in humans: perspectives of an epidemiologist. Semin Oncol (16) Grady G, Ernster VL. Endometrial cancer. In: Schottenfeld D, Fraumeni J Jr, editors. Cancer epidemiology and prevention. 2nd ed. New York (NY): (34) Horwitz RI, Feinstein AR. Alternative analytic methods for case-control studies of estrogens and endometrial cancer. N Engl J Med 1978;199: (17) Breslow NE, Day NE. Statistical methods in cancer research. Volume 1-The analysis of case-control studies. IARC Sci Publ 1980;32:192-278.
(35) Horwitz RI, Feinstein AR. Estrogens and endometrial cancer. Responses to (18) Fisher B, Dignam J, Bryant J, DeCillis A, Wickerham DL, Wolmark N, et arguments and current status of an epidemiologic controversy. Am J Med al. Five versus more than five years of tamoxifen therapy for breast cancer patients with negative lymph nodes and estrogen receptor-positive tumors.
J Natl Cancer Inst 1996;88:1529-42.
(19) Fromson JM, Sharp DS. The selective uptake of tamoxifen by human uterine tissue. J Obstet Gynaecol Br Commonw 1974;81:321-3.
1Editor's note: SEER is a set of geographically defined, population-based, (20) Groom GV, Griffiths K. Effect of the anti-oestrogen tamoxifen on plasma central cancer registries in the United States, operated by local nonprofit orga- levels of luteinizing hormone, follicle-stimulating hormone, prolactin, oes- nizations under contract to the National Cancer Institute (NCI). Registry data are tradiol and progesterone in normal pre-menopausal women. J Endocrinol submitted electronically without personal identifiers to the NCI on a biannual basis and the NCI makes the data available to the public for scientific research.
(21) Sherman BM, Chapler FK, Crickard K, Wycoff D. Endocrine conse- quences of continuous antiestrogen therapy with tamoxifen in premeno- Supported by a grant from the Wright Foundation (to L. Bernstein); by Public pausal women. J Clin Invest 1979;64:398-404.
Health Service grant CA17054 and by contracts N01PC67010, N01PC05229, (22) Lonning PE, Johannessen DC, Lien EA, Ekse D, Fotsis T, Adlercreutz H.
N01PC05230, and 01PC67006 from the National Cancer Institute, National Influence of tamoxifen on sex hormones, gonadotrophins and sex hormone Institutes of Health, Department of Health and Human Services; the Los Angeles binding globulin in postmenopausal breast cancer patients [published erra- County registry is also supported by subcontract 050-F-8709 from the California tum appears in J Steroid Biochem Mol Biol 1996;57:149]. J Steroid Bio- Public Health Foundation, which is supported by the California Department of (23) Zaino RJ, Satyasaroop PG, Mortel R. Hormonal therapy of human endo- We thank Dr. Salvacion LeBerthon, Courtney Mykytyn, Sarina Hieng, and metrial adenocarcinoma in a nude mouse model. Cancer Res 1985;45: Annette Lund (Los Angeles); Dr. Michelle West, Judy Anderson, Lori Odle, and Connie Mahoney (Iowa); Dr. Laurel Habel, Fran Chard, and Marjorie Blunt (24) Gorodeski GI, Barry R, Lunenfeld B, Geier A. Tamoxifen increases plasma (Seattle); and Judy Andrews (Atlanta) for their valuable contributions to this estrogen-binding equivalents and has an estradiol agonistic effect on his- research effort in the conduct of data collection activities.
tologically normal premenopausal and postmenopausal endometrium. Fertil The ideas and opinions expressed herein are those of the authors, and no endorsement by the State of California or the California Public Health Founda- (25) Boccardo F, Bruzzi P, Rubagotti A, Nicolo GU, Rosso R. Estrogen-like tion is intended or should be inferred.
action of tamoxifen on vaginal epithelium in breast cancer patients. On- Manuscript received February 17, 1999; revised July 19, 1999; accepted Au- Journal of the National Cancer Institute, Vol. 91, No. 19, October 6, 1999

Source: http://hydra.usc.edu/pm518b/literature/bernstein-etal99.pdf