Susan E. Hankinson, Walter C.Willett, JoAnn E. Manson, Graham A. Colditz, David J.Hunter, Donna Spiegelman,Robert L. Barbieri, Frank E.Speizer*


*Affiliations of authors: S. E. Hankinson, G. A. Colditz, D. J. Hunter, Channing Laboratory, Brigham and Women’s Hospital and Harvard Medi- cal School, and Department of Epidemiology, Har- vard School of Public Health, Boston, MA; W. C. Willett, Channing Laboratory, Brigham and Wom- en’s Hospital and Harvard Medical School, and De- partments of Epidemiology and Nutrition, Harvard School of Public Health; J. E. Manson, Channing Laboratory, Division of Preventive Medicine, De- partment of Medicine, Brigham and Women’s Hos- pital and Harvard Medical School, and Department of Epidemiology, Harvard School of Public Health; D. Spiegelman, Departments of Epidemiology and Biostatistics, Harvard School of Public Health; R. L. Barbieri, Department of Obstetrics and Gynecology, Brigham and Women’s Hospital and Harvard Medi- cal School; F. E. Speizer, Channing Laboratory, Brigham and Women’s Hospital and Harvard Medi- cal School, and Department of Environmental Health, Harvard School of Public Health.


Correspondence to: Susan E. Hankinson, Sc.D., Channing Laboratory, 181 Longwood Ave., Boston, MA 02115.


See ‘‘Notes’’ following ‘‘References.’’ © Oxford University Press


Background: A positive relationship has generally been observed between plasma estrogen levels and breast can- cer risk in postmenopausal women, but most of these studies have been small and few have evaluated specific estro- gen fractions (such as percent bioavail- able estradiol). In addition, few studies have evaluated plasma androgen levels in relation to breast cancer risk, and their results have been inconsistent. We prospectively evaluated relationships between sex steroid hormone levels in plasma and risk of breast cancer in postmenopausal women by use of a case-control study nested within the Nurses’ Health Study. Methods: Blood samples were collected during the pe- riod from 1989 through 1990. Among postmenopausal women not using hor- mone replacement therapy at blood collection (n = 11169 women), 156 women were diagnosed with breast cancer after blood collection but before June 1, 1994. Two control subjects were selected per case subject and matched with respect to age, meno- pausal status, month and time of day of blood collection, and fasting status at the time of blood collection. Results: From comparisons of highest and low- est (reference) quartiles, we observed statistically significant positive associa- tions with risk of breast cancer for cir- culating levels of estradiol (multivari- ate relative risk [RR] = 1.91; 95% confidence interval [CI] = 1.06-3.46), estrone (multivariate RR = 1.96; 95% CI = 1.05-3.65), estrone sulfate (multi- variate RR = 2.25; 95% CI = 1.23- 4.12), and dehydroepiandrosterone sul- fate (multivariate RR = 2.15; 95% CI = 1.11-4.17). We found no substantial as- sociations with percent free or percent bioavailable estradiol, androstenedi- one, testosterone, or dehydroepi- androsterone. The positive relation- ships were substantially stronger among women with no previous hor- mone replacement therapy. Conclu- sion: Our data, in conjunction with past epidemiologic and animal studies, provide strong evidence for a causal re- lationship between postmenopausal es- trogen levels and the risk of breast can- cer. [J Natl Cancer Inst 1998;90: 1292-9]


Substantial indirect evidence supports a central role for endogenous hormones in breast cancer development (1). Reproduc- tive factors such as early age at menarche, late age at menopause, and nulliparity are associated with an increased risk of breast cancer. The rate of increase in age- specific breast cancer incidence rates slows at menopause, a time when endog- enous estrogen levels decrease dramati- cally. In postmenopausal women, obesity (2) and use of postmenopausal hormone therapy (3), both positively related to plasma estrogen levels, also are positively related to breast cancer risk. Estrogens also induce mammary tumors in animals (4). Androgens may influence breast can- cer risk either directly (5) or indirectly, through their conversion to estradiol (6,7).


The relationships in postmenopausal women between hormone levels in plasma and the risk of breast cancer have been evaluated in six previous prospec- tive studies (8-13). For estrogens, the overall evidence supports a positive asso- ciation (14). However, in most studies, only one or two of the major circulating estrogens have been evaluated and, with one exception (8), the studies have been small, containing only 15-71 case sub- jects with breast cancer. For plasma an- drogens, the data are more limited and the results inconsistent.


To evaluate these relationships in de- tail, we conducted a prospective, nested case-control study within the Nurses’ Health Study cohort. We evaluated the levels of circulating estrogens and andro- gens in relation to the risk of breast can- cer. We also calculated estimates of effect that accounted for laboratory measure- ment error and the random within-person fluctuation in hormone levels over time (15).



Subjects and methods


Study Population


The Nurses’ Health Study cohort was established in 1976 when 121 700 female registered nurses, BO- 55 years of age, completed and returned a mailed questionnaire. The cohort continues to be followed every 2 years by questionnaire to update exposure status and to identify cases of newly diagnosed dis- ease. Data have been collected on many breast can- cer risk factors, including height, weight, age at menarche and menopause, age at birth of first child, parity, postmenopausal hormone use, diagnosis of benign breast disease, and family history of breast cancer.


During the period from 1989 through 1990, blood samples were collected from 32 826 cohort members (27% of the original cohort) who were 43-69 years of age when blood was collected. Details regarding the blood collection methods have been published (16). Briefly, each woman arranged to have her blood drawn and then shipped, via overnight courier and with an ice pack to keep the sample cool, to our laboratory, where it was processed and separated into plasma, red blood cell, and white blood cell components. Within 26 hours of being drawn, 97% of the samples were received in our laboratory. The stability of estrogens and androgens in whole blood for 24-48 hours has been documented previously (17). Since collection, samples have been archived at -130 °C or colder in continuously monitored liq- uid nitrogen freezers. As of 1994, the follow-up rate among women who gave a blood sample was 98%. The study was approved by the Committee on the Use of Human Subjects in Research at the Brigham and Women’s Hospital.


Both case and control subjects in this analysis are women who, at blood collection, were postmeno- pausal and had not used postmenopausal hormones for at least 3 months. The participants were defined as postmenopausal if they reported having a natural menopause or a bilateral oophorectomy. Women who reported a hysterectomy with either one or both ovaries remaining were defined as postmenopausal when they were 56 years old (if a nonsmoker) or 54 years old (if a current smoker), ages at which natural menopause had occurred in 90% of the respective cohorts.


Case subjects were women who had reported no cancer diagnosis before blood collection and who were diagnosed with breast cancer after blood col- lection but before June 1, 1994. Overall, 156 cases of breast cancer (140 invasive and 16 in situ) were reported from among the 11 169 women eligible at baseline. (The other 21 657 women were not eligible because they were premenopausal, were postmeno- pausal but were using postmenopausal replacement hormones, were of uncertain menopausal status, or had a prior cancer diagnosis.) All cases of breast cancer were confirmed by medical record review with one exception, in which the nurse confirmed the diagnosis but the medical record was unavail- able; because of the high confirmation rate (99%) upon medical record review, this case subject was kept in the analysis. The time from blood collection to diagnosis ranged from less than 1 month to 57 months (mean [standard deviation] = 28.7 [15.8] months). Two control subjects were matched per case subject by age (±2 years), month of blood col- lection, time of day that blood was drawn (±2 hours), and fasting status at the time of blood col- lection (≥10 hours since a meal versus <10 hours or unknown). Ninety-three percent of control matches were exact; the most relaxed match was within ±5 years of age, ±3 months of blood collection from case subjects, and ±7 hours for time of blood collection.


Laboratory Analyses


With the exception of estrone sulfate, all analyses were performed by the Nichols Institute (San Juan Capistrano, CA). Plasma samples were extracted with hexane-ethyl acetate (4:1, vol/vol), and the extract was applied to celite columns (celite in eth- ylene glycol). The steroids were eluted from the col- umns in the following fractions: fraction 1, 3.5 mL of iso-octane (androstenedione); fraction 2, 3.5 mL of iso-octane containing 10% ethyl acetate (dehy- droepiandrosterone [DHEA] and testosterone); frac- tion 3, 3.0 mL of iso-octane containing 15% ethyl acetate (estrone); and fraction 4, 5.0 mL of iso- octane containing 40% ethyl acetate (estradiol). Fractions 1-4 were then assayed by radioimmuno- assay (18—21). Dehydroepiandrosterone sulfate (DHEAS) was assayed by radioimmunoassay with- out a prior separation step (22). Percent free estra- diol (i.e., percent nonprotein bound) was assayed by use of equilibrium dialysis (23,24); the percent dia- lyzable estradiol was calculated as described by Ver- muelen et al. (24). The percent bioavailable estradiol (i.e., percent free plus percent albumin-bound estra- diol) was assayed by use of an ammonium sulfate precipitation (25,26). All case-control-control trip- let samples were assayed together; the samples were ordered randomly within a triplet and labeled so that the laboratory could not identify the case-control status. Although all members of a triplet were ana- lyzed at the same time, the triplets were analyzed in up to three different batches (sent in 1992, 1993, and 1996).


For estrone sulfate, the first two batches of samples were assayed at the laboratory of Dr. C. Longcope at the University of Massachusetts Medi- cal Center, Worcester, and the third batch was as- sayed at the Nichols Institute. In each laboratory, after extraction of estrone, estrone sulfate was as- sayed by radioimmunoassay of estrone, after en- zyme hydrolysis, organic extraction, and separation by column chromatography (27).


In each batch of samples, we interspersed plasma replicates (one replicate per 10 case and/or control samples) that were labeled to preclude their identi- fication by the assaying laboratory; these replicate samples were used to assess laboratory precision. Within-batch laboratory coefficients of variation ranged from 6% (percent bioavailable estradiol) to 13.6% (DHEA).


The assay detection limit was 2 pg/mL for estra- diol, 0.5% for both percent free estradiol and percent bioavailable estradiol, 10 pg/mL for estrone, 50 pg/ mL for estrone sulfate (in each laboratory), 3 ng/dL for androstenedione, 1 ng/dL for testosterone, 3 ng/ dL for DHEA, and 5 μg/dL for DHEAS. When plasma hormone values were reported as less than the detection limit, we set the value to halfthis limit (which occurred only for estrone [n = 6], estrone sulfate [n = 2], and DHEAS [n = 2]).


Reproducibility Study


Three hundred ninety Nurses’ Health Study par- ticipants who gave a first blood sample during the period from 1989 through 1990 were asked to pro- vide two additional samples that were collected dur- ing the following 2 years. The women were post- menopausal, had not used postmenopausal hormones for at least 3 months, and had no previous diagnosis of cancer (except nonmelanoma skin can- cer); these criteria were applied at each sample col- lection. Of the 390 women, 186 (48%) sent two additional samples. A random sample of 80 of these women who had all three samples drawn between 6 AM and 12 noon was sent for hormone analysis, at the same laboratories used for the main study, and forms the basis of the reproducibility study. Addi- tional details regarding this study are provided else- where (15).


Data Analyses


We used quartile categories, with cut points based on the distribution in the control subjects, for the purpose of summarizing breast cancer risk according to plasma hormone level. For most of the hormones, the mean and standard deviation of both the control values and the quality-control replicates were very similar across batches; thus, quartile cut points were made according to the distribution in the control subjects overall. The lowest quartile was used as the referent in all analyses. For estrone, estrone sulfate, and DHEA, the median value for the control subjects varied by as much as 40% between batches, so that quartile cut points based on all control subjects com- bined resulted in uneven batch-specific distributions (e.g., the lowest quartile of estrone contained 12% of the control subjects from the first two batches but 41% of the control subjects from the third batch). Because the mean value of the quality-control rep- licates in each of the datasets varied in the same manner for these three assays, much (if not all) of this difference appeared due to laboratory drift rather than to true differences in hormone levels be- tween the batches. Thus, for these three hormones, we defined batch-specific quartile cut points. In ad- dition, in all analyses, we controlled for batch. For several hormones (e.g., estradiol), the control distri- bution was unequal across quartiles because of mul- tiple identical hormone values.


One matched set was removed from the analysis because the case subject’s estrogen values were in the premenopausal range (estradiol = 411 pg/mL). Individual values more than 2.5-fold higher than the normal range according to the assaying laboratory also were removed; this resulted in the removal of two testosterone values only. In addition, several women did not have a sufficient volume of plasma for all assays. The final number of case and control samples available for each hormone analysis is pro- vided in Table 1.


Table 1. . Median and range* of plasma hormone levels for case and matched control subjects.

KmH5h1gS1WeuLvohsBe5Azylw226w5tlzQ.png



To test for differences in hormone levels between case and control subjects, we used mixed-effects re- gression models for clustered data to adjust for pos- sible confounding due to the matching factors and for any residual correlation between case and control subjects within the matched set (28). To compare proportions between case and control subjects, we used the Mantel-Haenszel test (29). We used con- ditional logistic regression analyses to estimate rela- tive risks (RRs) (odds ratios) and 95% confidence intervals (CIs) (30). In analyses stratified by prior postmenopausal hormone use, however, we used un- conditional logistic regression, controlling for the matching factors, to maximize our sample size. We conducted tests for trend by modeling the natural logarithm of the hormone level as a continuous vari- able and calculating a Wald statistic (31). All P val- ues are based on two-sided tests. The regression cali- bration method was used to correct RRs and 95% CIs for laboratory measurement error and random within-person variability (32—35). The within- person variance was calculated from the reproduci- bility study and the between-person variance from the current case-control study. (Thus, intraclass cor- relation coefficients are slightly different from the previously published values.) In these analyses, hor- mone levels were log transformed to lessen the in- fluence of a small number of high or low values. Because the measurement error correction methods require that the relationship between disease and ex- posure be linear on a logistic scale, restricted cubic spline models (36) for breast cancer incidence in relation to each log-transformed hormone value were fit to the data. With this technique, as well as formal significance testing criteria for nonlinearity, with just one exception (DHEA), none of the hor- mones showed substantial evidence of departure from a linear relation on the logarithmic RR scale. DHEA was modeled on its original scale.



Results


The women in this analysis ranged in age from 46 years to 69 years (mean age = 62 years) and had been menopausal for at least 1 year and up to 40 years (mean = 12 years). Compared with control sub- jects, case subjects had an earlier mean age at menarche (12.4 years versus 12.7 years) and a later mean age at the birth of their first child (26.0 years versus 25.3 years) and were more likely to have re- ported a family history of breast cancer (19% versus 15%), although none of these differences were statistically significant. We observed that case subjects, when compared with control subjects, had sig- nificantly higher plasma levels of estra- diol, estrone, estrone sulfate, testosterone, and DHEAS but no substantial difference in levels of the other steroid hormones (Table 1).


In the simple conditional models, women in the top quartile of plasma es- trone and estrone sulfate levels had an ap- proximately twofold increase in breast cancer risk, which was statistically sig- nificant (for estrone, RR = 1.77 [95% CI = 1.01-3.11]; for estrone sulfate, RR = 2.12 [95% CI = 1.21-3.71]). For DHEAS, women in the top 75% of levels appeared to have an increase in breast cancer risk compared with women with the lowest levels. Modest, and generally nonsignificant, positive associations were noted for percent free estradiol, andro- stenedione, and testosterone and breast cancer risk. We observed little association with either percent bioavailable estradiol or DHEA. When we evaluated absolute levels of free and bioavailable estradiol, the associations were similar to those for total estradiol.


When a number of established breast cancer risk factors were controlled for sta- tistically (Table 2), the relationships tended to strengthen somewhat, primarily because of control for age at birth of first child and body mass index at age 18 years. The association with estradiol was statistically significant (RR = 1.91; 95% CI = 1.06-3.46). Body mass index at age 18 was included in these models because it is inversely related to postmenopausal breast cancer risk (2); thus, we expected it could be a confounder. In contrast, when we included body mass index at the time of blood collection in each of the models, RRs for the estrogens were modestly at- tenuated, because, in postmenopausal women, body mass index is a major determinant of estrogen levels (16). For example, when the top quartile is compared with the bottom quartile, the RR decreased from 1.91 to 1.69 (95% CI = 0.83-3.42) for estradiol and from 1.96 to 1.75 (95% CI = 0.90-3.38) for estrone.


Table 2. Relative risk (RR) of breast cancer (and 95% confidence intervals [CI]) by category of plasma hormone levels among postmenopausal women in the Nurses’ Health Study.

C9_6IXX9V-wDkKc9HqSi4ls0MuZa3wHUvg.png



When we assessed the relationships between plasma hormones and the risk of breast cancer after excluding in situ breast cancer cases (n = 16), we observed nearly identical RRs. We also evaluated these relationships after excluding data from the 30 breast cancer cases that had been diagnosed within 1 year of blood collection, to assess whether the positive associations might be due to an influence of the breast cancer itself on hormone lev- els. With the exception of percent free and percent bioavailable estradiol, where the relationships were slightly strengthened (comparison of the top quartile with the bottom quartile, 1.69 [95% CI = 0.86- 3.32] and 1.50 [95% CI = 0.79-2.84], respectively), results again did not differ materially.


We next evaluated the relationships between hormone levels and the risk of breast cancer according to postmeno- pausal hormone use before blood collec- tion (i.e., never versus past use) (Table 3). We hypothesized that our single hormone measure would best reflect long-term en- dogenous hormone exposure among the never users and, therefore, we might see stronger associations in this group. Be- cause of the small number of cases in each of the groups, we included in the statistical models only the matching fac- tors and other most important covariates (Table 3). Among those who had never used postmenopausal hormones, the rela- tionships with the estrogens, particularly estradiol and estrone sulfate, were mark- edly strengthened (comparison of the top quartile with the bottom quartile: for es- tradiol, RR = 3.53 [95% CI = 1.55- 8.03]; for estrone sulfate, RR = 4.34 [95% CI = 1.87-10.1]). The association with DHEAS also was stronger. In con- trast, the relationships among past hor- mone users were weak (or null) and not statistically significant, although the 95% CIs were wide.


Table 3. Multivariate relative risk* of breast cancer by plasma hormone level, according to use of postmenopausal hormones before blood collection.

5FpMDvPJ7DHgjHpk1_FSw0i1dvamo2lv5g.png



Most of the steroid hormones are posi- tively correlated. For example, the Spear- man correlations for estradiol with es- trone, testosterone, and DHEAS were .67, .45, and .27, respectively. Therefore, we evaluated the independent association of each of the hormones with breast cancer risk, among all case and control subjects combined, when estradiol also was in- cluded in the statistical model. The RRs for testosterone were substantially attenu- ated (comparison of the top quartile with the bottom quartile: RR = 1.08 [95% CI = 0.52-2.25]), whereas the RRs for es- trone (RR = 1.50 [95% CI = 0.64- 3.54]), DHEAS (RR = 1.90 [95% CI = 0.96-3.77]), and estradiol itself were only modestly reduced. When estrone and es- trone sulfate were included in the same statistical model, neither was attenuated, although the 95% CIs for each widened considerably.


We next corrected the associations for laboratory error and random within- person variability; in these analyses, hor- mone levels were modeled as continuous variables (Table 4). The RR (based on a contrast in hormone levels from the 12.5 to the 87.5 percentiles of the distribution, corresponding to the medians of the bot- tom quartile and the top quartile, respec- tively, as shown in Table 1) for estradiol strengthened considerably, increasing from 1.77 to 2.42. Similarly, the relation- ships with each of the other hormones strengthened somewhat, although only the relationships with estrone, estrone sulfate, percent free estradiol, DHEAS, and tes- tosterone were statistically significant. As in the categorical analyses, the association with testosterone was substantially at- tenuated after we controlled for estradiol.


Table 4. Correction of multivariate relative risk (RR)* estimates and 95% confidence intervals (CIs) for random within-person measurement error.

vJm_TDFlLWizP3XssjeZ3weQDX14nQwcVQ.png




Discussion


We observed positive associations be- tween circulating levels of estradiol, es- trone, estrone sulfate, and DHEAS and risk of breast cancer in postmenopausal women. In contrast, we found no substantial asso- ciations for percent bioavailable estradiol, androstenedione, or DHEA in relation to breast cancer. The positive relationships were considerably stronger among women with no previous use of hormone replace ment therapy after menopause.


Strengths of our study include that it was prospective and relatively large. In addition, we were able to evaluate nine steroid hormones or hormone fractions, all of which were assayed with good pre- cision. By using multiple hormone mea- sures from a subset of study participants, we were able to correct our RR estimates for the random (and largely biologic) variation in hormone levels that cannot ordinarily be captured by a single hor- mone measurement.


Evidence from our study, in conjunc-tion with that from other recent prospec- tive studies (8-12), supports a strong pre- dictive role for plasma estradiol levels in relation to breast cancer risk among post- menopausal women. In only one small prospective study (13) has a positive as- sociation not been observed. Although considerably larger RRs have been re- ported for contrasts in levels generally similar to ours (11,12), these two studies had sample sizes of only 24 and 61 case subjects, respectively; thus, their confi- dence limits broadly overlap ours. More- over, some of the heterogeneity in RRs between studies may be due to various prevalences of past postmenopausal hor- mone use in study subjects. The magni- tude of the associations also might be ex- pected to vary because of different sensitivities and specificities of the labo- ratory assays used in the studies (37,38); this limitation makes the comparison of results between studies difficult and esti- mation of the increase in disease risk per unit increase in estradiol levels (as is done with plasma cholesterol level and heart disease risk) currently infeasible.


Free estradiol or bioavailable estradiol is hypothesized to be readily available to the breast tissue and thus is considered to be the most biologically active estrogen fraction(s) (39). As such, compared with total estradiol, a stronger relationship be- tween one of these fractions and breast cancer risk might be expected. However, the epidemiologic evidence has not been consistent (8,9,40-43). We noted only a marginally significant positive relation- ship with percent free estradiol. We also observed no substantial relationship be- tween percent bioavailable estradiol and risk, in contrast to the only previous large prospective study of this issue (compari- son of the top quartile with the bottom quartile, RR = 4.4) (8). These differences seem unlikely to be due to confounding or to different levels of measurement error. Our laboratory coefficient of variation was small, and measurement error correc- tion did not increase the estimates appre- ciably. We previously documented that our blood collection methods did not alter levels of percent free estradiol (17), sug- gesting that a change in the bioavailable fraction also is unlikely. The average age (59 years versus 62 years) and lengths of follow-up times (5 years versus 2.5 years) of the populations in the study by Toniolo et al. (8) and in our study also were simi- lar. In addition, although the percent bio- available estradiol values varied substan- tially between the two studies, the results of the two assays are highly correlated. We sent 112 of our control samples for analysis to the laboratory used by Toniolo et al.; the Spearman correlation between the percent bioavailable estradiol assays from the two laboratories was .91. To our knowledge, these estrogen fractions have not been evaluated in any other large pro- spective studies; thus, additional assess- ments are needed.


Estrone sulfate is the most abundant circulating estrogen in postmenopausal women (44,45) and a major component of some postmenopausal hormone prepara- tions. Although Dorgan et al. (10), in the only other prospective study to examine this hormone, observed little association with breast cancer risk, they were unable to rule out an approximately twofold in- crease in risk, such as we observed among women in the top 25% of the distribution compared with those in lower exposure categories.


Androgens have been hypothesized to increase breast cancer risk either directly by increasing the growth and proliferation of breast cancer cells (5) or indirectly by their conversion to estrogen (6,7). Testos- terone has been positively associated with breast cancer in most (10-12,46-49) but not all (50,51) previous studies. However, the positive association has tended to weaken after controlling for total estradiol (or another estrogen fraction) (12,46), similar to our findings, suggesting that in- creased levels of testosterone may have a modest, but indirect, association with breast cancer through its conversion to es- tradiol.


DHEA and DHEAS are adrenal andro-gens that decrease substantially with in- creasing age and have little documented physiologic role (52). DHEA adminis- tered to rodents can decrease the risk of spontaneous and chemically induced can- cers (53). However, in postmenopausal women, DHEA has been proposed to act like an estrogen in stimulating cell growth (52), in part because of the estrogenic ef- fect of its major metabolite, 5-andro- stenediol (54).


DHEAS has been evaluated in relation to breast cancer risk in five previous prospective studies; with one exception (55) (21 case subjects), nonsignificant positive associations have been reported (10,11,46,56), although in one of these studies (46) the weak positive association became inverse after controlling for estra- diol. We observed a positive association that was essentially independent of estra- diol. In the two previous assessments of DHEA and breast cancer (10,56), a statis- tically significant positive association was observed. We found no statistically sig- nificant association but cannot rule out a modest positive relationship. As a whole, these findings should serve to caution against the increasing use of pharmaco- logic doses of DHEA as an ‘‘anti-aging’’ agent. DHEA and DHEAS are metaboli- cally interconvertible, and after oral ad- ministration of DHEA, circulating levels of DHEAS rise substantially (57). Cer- tainly, epidemiologic evidence does not support a decreased risk of breast cancer with increasing levels of these androgens and, in fact, suggests a possible positive association. In addition, DHEA supple- mentation may increase levels of plasma insulin-like growth factor-I (58), a hor- mone that has recently been associated with risk of breast cancer (59,60) and prostate cancer (61).


Estrogen (and some androgen) levels in normal breast tissue are generally much higher than levels in plasma, and levels in malignant tissue are higher than those in normal breast tissue (62-64). These dif- ferences may be due to enzyme activities in normal and malignant breast cells that result in the local conversion of andro- gens to estrogens, estrone sulfate to es- trone, and estrone to estradiol (6,63,65). Although several reports (62-64,66,67) have indicated that there is little if any correlation between plasma and tissue ste- roid levels, these studies were all small (n = ≤ 14 women) and the correlations were not provided. Given our findings and those of others described above, it seems unlikely that these levels are entirely un- correlated. A low correlation would sug- gest, however, that the relationships be- tween tissue hormone levels and breast cancer risk may be stronger than those observed with our plasma surrogates.


Our data, in conjunction with past epi- demiologic (1-3,8-12) and animal (4) studies, provide strong evidence for a cau- sal relationship between postmenopausal plasma estrogen levels and risk of breast cancer (68). However, additional studies are needed before conclusions can be made as to whether total estradiol or other specific fractions are most important to risk. Testosterone most likely has a mod- est, indirect influence on risk through its conversion to estradiol, and increasing evidence suggests a positive relationship between DHEAS and the risk of breast cancer. Although higher estrogen levels may have both beneficial (69) and ad- verse effects, reducing the levels or activ- ity of endogenous estrogens may be a promising means for preventing breast cancer in postmenopausal women.



Notes


Supported by Public Health Service grants CA40356 and CA49449 from the National Cancer Institute, National Institutes of Health, Department of Health and Human Services. S. Hankinson was partially supported by a career development award DAMD17-96-1-6021 from the U.S. Army Medical Research and Materiel Command.


Health Study for their continuing dedication and commitment and to Dr. Dominique Michaud, Rachel Meyer, Michele Lachance, Kathryn Starzyk, and San- dra Melanson for their expert and unfailing assistance.


We also thank Dr. C. Longcope, whose laboratory conducted a number of the estrone sulfate assays.


Manuscript received February 10, 1998; revised June 1, 1998; accepted June 29, 1998.




References


Bernstein L, Ross RK. Endogenous hormones and breast cancer risk. Epidemiol Rev 1993;15:48-65.  

Huang Z, Hankinson SE, Colditz GA, Stampfer MJ, Hunter DJ, Manson JE, et al. Dual effects of weight and weight gain on breast cancer risk. JAMA 1997;278:1407-11. 

Collaborative Group on Hormonal Factors in 

Breast Cancer. Breast cancer and hormone replacement therapy: collaborative reanalysis of data from 51 epidemiological studies of52,705 women with breast cancer and 108,411 women without breast cancer. Lancet 1997;350:1047-59. 

Dao TL. The role of ovarian steroid hormones in mammary carcinogenesis. In: Pike MC, Sii-teri PK, Welsch CW, editors. Hormones and breast cancer. Banbury Report No. 8. Cold Spring Harbor (NY): Cold Spring Harbor Laboratory; 1981. p. 281-95. 

Lippman M, Bolan G, Huff K. The effects of androgens and antiandrogens on hormone-responsive human breast cancer in long-term tissue culture. Cancer Res 1976;36:4610-8. 

James VH, McNeill JM, Lai LC, Newton CJ, Ghilchik MW, Reed MJ. Aromatase activity in normal breast and breast tumor tissues: in vivo and in vitro studies. Steroids 1987;50:269-79. 

Thijssen JH, Blankenstein MA, Donker GH, Daroszewski J. Endogenous steroid hormones and local aromatase activity in the breast. J. Steroid Biochem Mol Biol 1991;39:799-804. 

Toniolo PG, Levitz M, Zeleniuch-Jacquotte A, Banerjee S, Koenig KL, Shore RE, et al. Aprospective study of endogenous estrogens and breast cancer in postmenopausal women. J Natl Cancer Inst 1995;87:190-7. 

Helzlsouer KJ, Alberg AJ, Bush TL, Longcope C, Gordon GB, Comstock GW. A prospective study of endogenous hormones and breast can-cer. Cancer Detect Prev 1994;18:79-85. 

Dorgan JF, Longcope C, Stephenson HE Jr, Falk RT, Miller R, Franz C, et al. Relation of prediagnostic serum estrogen and androgen levels to breast cancer risk. Cancer Epidemiol Biomarkers Prev 1996;5:533-9. 

Berrino F, Muti P, Micheli A, Bolelli G, Krogh V, Sciajno R, et al. Serum sex hormone levels after menopause and subsequent breast cancer. J. Natl Cancer Inst 1996;88:291-6. 

Thomas HV, Key TJ, Allen DS, Moor JW, Dowsett M, Fentiman IS, et al. A prospective study of endogenous serum hormone concentrations and breast cancer risk in post menopausal women on the island of Guernsey. Br. J. Cancer 1997;76:401-5. 

Garland CF, Friedlander NJ, Barrett-Connor E, Khaw KT. Sex hormones and postmenopausal breast cancer: a prospective study in an adult community. Am J Epidemiol 1992;135:1220-30. 

Thomas HV, Reeves GK, Key TJ. Endogenous estrogen and postmenopausal breast cancer: a quantitative review. Cancer Causes Control 1997;8:922-8. 

Hankinson SE, Manson JE, Spiegelman D, Willett WC, Longcope C, Speizer FE. Reproducibility of plasma hormone levels in post menopausal women over a 2-3-year period. Cancer Epidemiol Biomarkers Prev 1995;4:649-54. 

Hankinson SE, Willett WC, Manson JE, Hunter DJ, Colditz GA, Stampfer MJ, et al. Alcohol, height, and adiposity in relation to estrogen and prolactin levels in postmenopausal women. J. Natl Cancer Inst 1995;87:1297-302. 

Hankinson SE, London SJ, Chute CG, Barbieri RL, Jones L, Kaplan LA, et al. Effect of transport conditions on the stability of biochemical markers in blood. Clin Chem 1989;35:2313-6. 

Abraham GE, Tulchinsky D, Korenman SG. Chromatographic purification of estradiol-17ß for use in radio-ligand assay. Biochem Med 1970;3:365-8. 

Mikhail G, Chung HW. Radioimmunoassay of plasma estrogens. Use of polymerized antibodies. In: Peron FG, Caldwell BV, editors. Immunologic methods in steroid determination. New York: Appleton-Century-Crofts; 1970. p.113. 

Kinouchi T, Pages L, Horton R. A specific radioimmunoassay for testosterone in peripheral plasma. J Lab Clin Med 1973;82:309-16. McNatty KP, Smith DM, Makris A, Osathanondh R, Ryan KJ. The microenvironment ofthe human antral follicle: interrelationships among the steroid levels in antral fluid, the population of granulosa cells, and the status of the oocyte in vivo and in vitro. J Clin Endocrinol Metab 1979;49:851-60. 

Buster JE, Abraham GE. Radioimmunoassay of plasma dehydroepiandrosterone sulfate. Anal Lett 1972;5:543. 

Moll GW Jr, Rosenfield RL, Helke JH. Estradiol-testosterone binding interactions and free plasma estradiol under physiological conditions. J Clin Endocrinol Metab 1981;52:868-74. 

Vermeulen A, Stoica T, Verdonck L. The apparent free testosterone concentration, an index of androgenicity. J Clin Endocrinol Metab 1971;33:759-67. 

Sodergard R, Backstrom T, Shanbhag V, Carstensen H. Calculation of free and bound fractions of testosterone and estradiol-17β to human plasma proteins at body temperature. J. Steroid Biochem 1982;16:801-10. 

Stumpf PG, Nakamura RM, Mishell DR Jr. Changes in physiologically free circulating estradiol and testosterone during exposure to levonorgestrel. J Clin Endocrinol Metab 1981;52:138-43. 

Franz C, Watson D, Longcope C. Estrone sulfate and dehydroepiandrosterone sulfate concentrations in normal subjects and men with cirrhosis. Steroids 1979;34:563-73. 

Liang IK, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986;73:13-22. 

Rothman KJ. Modern epidemiology. 1st ed. Boston: Little, Brown & Co.; 1986. p. 237-83. 

Rosner B. Fundamentals of biostatistics. 4th ed. Belmont (CA): Wadsworth Publishing Co.; 1993. 

Hosmer DW, Lemeshow S. Applied logistic regression. New York: Wiley; 1989. 

Carroll RJ, Ruppert D, Stefanski LA. Measurement error in nonlinear models. London: Chapman & Hall; 1995. 

Prentice RL. Covariate measurement errors and parameter estimation in a failure time regression model. Biometrika 1982;69:331-42. 

Wang CY, Hsu L, Feng ZD, Prentice RL. Regression calibration in failure time regression. Biometrics 1997;53:131-45. 

Spiegelman D, McDermott A, Rosner B. Regression calibration methods for correcting measurement-error bias in nutritional epidemiology. Am J. Clin Nutr 1997;65(4 Suppl):1179S-1186S. 

Durrleman S, Simon R. Flexible regression models with cubic splines. Stat Med 1989;8:551-61. 

Hankinson SE, Manson JE, London SJ, Willett WC, Speizer FE. Laboratory reproducibility of endogenous hormone levels in postmenopausal women. Cancer Epidemiol Biomarkers Prev 1993;3:51-6. 

Potischman N, Falk RT Laiming VA, Siiteri PK, Hoover RN. Reproducibility of laboratory assays for steroid hormones and sex hormone binding globulin. Cancer Res 1994;54:5363-7. 

Siiteri PK, Hammond GL, Nisker JA. Increased availability of serum estrogens in breast cancer: a new hypothesis. Banbury Rep 1981;8:87-106. 

Moore JW, Clark GM, Hoare SA, Millis RR, Hayward JL, Quinlan MK, et al. Binding of oestradiol to blood proteins and aetiology of breast cancer. Int J Cancer 1986;38:625-30. 

Siiteri PK, Simberg N, Murai J. Estrogens and breast cancer. Ann N Y Acad Sci 1986;464:100-5. 

Reed MJ, Cheng RW, Noel CT, Dudley HA, James VH. Plasma levels of estrone, estrone sulfate, and estradiol and the percentage of un-bound estradiol in postmenopausal women with and without breast disease. Cancer Res 1983;43:3940-3. 

Ernster VL, Wrensch MR, Petrakis NL, King EB, Miike R, Murai J, et al. Benign and ma lignant breast disease: initial study results of serum and breast fluid analyses of endogenous estrogens. J Natl Cancer Inst 1987;79:949-60. 

Longcope C, Franz C, Morello C, Baker R, Johnston CC Jr. Steroid and gonadotropin levels in women during the peri-menopausal years. Maturitas 1986;8:189-96. 

Roberts KD, Rochefort JG, Bleau G, Chapdelaine A. Plasma estrone sulfate levels in postmenopausal women. Steroids 1980;35:179-87. 

Zeleniuch-Jacquotte A, Bruning PF, Bonfrer JM, Koenig KL, Shore RE, Kim MY, et al. Relation of serum levels of testosterone and dehydroepiandrosterone sulfate to risk of breast cancer in postmenopausal women. Am J. Epidemiol 1997;145:1030-8. 

Secreto G, Toniolo P, Berrino F, Recchione C, Cavalleri A, Pisani P, et al. Serum and urinary androgens and risk of breast cancer in postmenopausal women. Cancer Res 1991;51:2572-6. 

Adami HO, Johansson ED, Vegelius J, Victor A. Serum concentrations of estrone, androstenedione, testosterone and sex-hormone-binding globulin in postmenopausal women with breast cancer and in age-matched controls. Ups J Med Sci 1979;84:259-74. 

Secreto G, Recchione C, Cavalleri A, Miraglia M, Dati V. Circulating levels of testosterone, β-oestradiol, luteinising hormone and prolactin in postmenopausal breast cancer patients. Br J Cancer 1983;47:269-75. 

Wysowski DK, Comstock GW, Helsing KJ, Lau HL. Sex hormone levels in serum in relation to the development of breast cancer. Am J. Epidemiol 1987;125:791-9. 

Garland CF, Friedlander NJ, Barrett-Connor E, Khaw KT. Sex hormones and postmenopausal breast cancer: a prospective study in an adult community. Am J Epidemiol 1992;135:1220-30. 

Ebeling P, Koivisto VA. Physiological importance of dehydroepiandrosterone. Lancet 1994;343:1479-81. 

Gordon GB, Shantz LM, Talalay P. Modulation of growth, differentiation and carcinogenesis by dehydroepiandrosterone. Adv Enzyme Regul 1987;26:355-82. 

Seymour-Munn K, Adams J. Estrogenic effects of 5-androstene-3β, β-diol at physiological concentrations and its possible implication in the etiology of breast cancer. Endocrinology 1983;112:486-91. 

Barrett-Connor E, Friedlander NJ, Khaw KT. Dehydroepiandrosterone sulfate and breast cancer risk. Cancer Res 1990;50:6571-4.  

Gordon GB, Bush TL, Helzlsouer KJ, Miller SR, Comstock GW. Relationship of serum levels of dehydroepiandrosterone and dehydroepi-androsterone sulfate to the risk of developing postmenopausal breast cancer. Cancer Res 1990;50:3859-62. 

Longcope C. Metabolism of dehydroepi androsterone. Ann N Y Acad Sci 1995;774: 143-8. 

Morales AJ, Nolan JJ, Nelson JC, Yen SS. Effects of replacement dose of dehydroepiandrosterone in men and women of advancing age published erratum appears in J Clin Endocrinol Metab 1995;80:2799]. J Clin Endocrinol Metab 1994;78:1360-7. 

Bruning PF, Van Doorn J, Bonfrer JM, Van Noord PA, Korse CM, Linders TC, et al. Insulin-like growth-factor-binding protein 3 is decreased in early-stage operable premenopausal breast cancer [published erratum appears in Int J Cancer 1995;63:762]. Int J Cancer 1995;62: 

266-70. 

Hankinson SE, Willett WC, Colditz GA, Hunter DJ, Michaud DS, Deroo B, et al. Cir culating concentrations of insulin-like growth 

factor-I and risk of breast cancer. Lancet 1998; 351:1393-6. 

Chan JM, Stampfer MJ, Giovannucci E, Gann PH, Ma J, Wilkinson P, et al. Plasma insulin-like growth factor-I and prostate cancer risk: a prospective study. Science 1998;279:563-6. 

Thijssen JH, Blankenstein MA. Endogenous oestrogens and androgens in normal and malignant endometrial and mammary tissues. Eur J. Cancer Clin Oncol 1989;25:1953-9. 

Vermeulen A, Deslypere JP, Paridaens R, Leclercq G, Roy F, Heuson JC. Aromatase, 17β-hydroxysteriod dehydrogenase and intratissular sex hormone concentrations in cancerous and normal glandular breast tissue in post menopausal women. Eur J Cancer Clin Oncol 1986;22:515-25. 

van Landeghem AA, Poortman J, Nabuurs M, Thijssen JH. Endogenous concentration and subcellular distribution of estrogens in normal and malignant human breast tissue. Cancer Res 1985;45:2900-6. 

Perel E, Blackstein ME, Killinger DW. Aromatase in human breast carcinoma. Cancer Res 1982;42(8 Suppl):3369s-3372s. 

Iqbal MJ, Valyani SH, Roberts JV. Mechanisms of sex-steroid accumulation in the aetiology of human breast carcinoma: the role of intracellular sex hormone binding globulin. Anticancer Res 1988;8:479-81. 

Thijssen JH, Blankenstein MA, Miller WR, Milewicz A. Estrogens in tissues: uptake from the peripheral circulation or local production. Steroids 1987;50:297-306. 

Colditz GA. Relationship between estrogen levels, use of hormone replacement therapy, and breast cancer. J Natl Cancer Inst 1998;90: 814-23. 

Cauley JA, Gutai JP, Kuller LH, LeDonne D, Sandler RB, Sashin D, et al. Endogenous estrogen levels and calcium intakes in postmenopausal women. Relationships with cortical bone measures. JAMA 1988;260:3150-5.