Quantifying estrogen metabolism: an evaluation of the reproducibility and validity of enzyme immunoassays for 2-hydroxyestrone and 16alpha-hydroxyestrone in urine.

Rapid and simple enzyme immunoassays (EIAs) were recently developed to measure 2-hydroxyestrone and 16alpha-hydroxyestrone in unextracted urine. The balance between these competing estrogen metabolism pathways may serve as a biomarker of breast cancer risk. Before testing these assays in epidemiologic studies, we evaluated their reproducibility, and validity relative to gas chromatography-mass spectroscopy (GC-MS). Overnight 12-hr urine collections from five midfollicular premenopausal women, five midluteal premenopausal women, and five postmenopausal women were aliquoted and stored at -70 degrees C. Two aliquots from each woman were assayed with the EIAs in a random, blinded order, monthly over 4 months and 1 year later. Reproducibility over 4 months was good for both metabolites in premenopausal women (coefficient of variation = 8-14%) and satisfactory in postmenopausal women (approximately 19%). Reproducibility over 12 months remained good in premenopausal women, but was poor in postmenopausal women, with mean readings increasing 50 to 100%. Wide variation in estrogen metabolite levels enabled a single EIA measurement to characterize individual differences among premenopausal women in midfollicular (intraclass correlation coefficient = 98-99%) and midluteal phase (85-91%). A narrower range in metabolite levels among postmenopausal women reduced discrimination (78-82%). The correlation between EIA and GC-MS measurement was excellent for both metabolites (r>0.9), except for 2-hydroxyestrone in postmenopausal women (r=0.6). Analysis of absolute agreement suggested that both EIAs were less sensitive than GC-MS, and each detected nonspecific background. The low concentration of estrogen metabolites in urine from postmenopausal women may explain the problems with reproducibility and validity in this menstrual group. Accordingly, more sensitive EIAs have been developed and are now being evaluated.


Introduction
Experimental, epidemiologic, and clinical evidence strongly suggests that endogenous estrogens influence breast carcinogenesis although the active form(s) of estrogen and the specific mechanisms remain unclear (1). In 1982, the relative importance of two major, mutually exclusive pathways for estradiol oxidation, 2-hydroxylation and 16a-hydroxylation, was postulated to determine a woman's risk of breast cancer, based on the finding that estrogen 16a-hydroxylase activity was increased in women with breast cancer (2). Over time, the ratio of 2-hydroxyestrone to 16a-hydroxyestrone (the 2/16 ratio) has been emphasized as an indicator of the balance between the two pathways and of reduced breast cancer risk. Estradiol is first reversibly converted to estrone by 17,B oxidation; most of the estrone is then irreversibly oxidized to 2-hydroxyestrone or 16a-hydroxyestrone, the initial metabolites formed along these two pathways. Recently, alteration of the balance between these two metabolic pathways has been proposed as the mechanism by which certain pesticides, herbicides, plastics, and other xenoestrogens (foreign estrogens) may increase the risk of breast cancer (3,4) and by which a low-fat diet (5) or indoles (6,7) may decrease the risk.
In vitro studies, using cell cultures and organ explants, and animal studies indicate that 16a-hydroxyestrone is a potent estrogen, genotoxic, and tumorigenic and that 2-hydroxyestrone is a weak estrogen and an estrogen antagonist (8,9). Evidence in humans is more limited. Breast tissue from breast cancer patients had nearly 5 times more 166a-hydroxylase activity than comparable tissue from women without cancer (10). A metabolic study in humans demonstrated that a decrease in dietary fat decreased 16a-hydroxylated estrogens in the urine (5), although studies of urinary estrogen profiles in vegetarian and omnivorous women and in women with breast cancer indicated the relationships are complex (11)(12)(13). To our knowledge, the relationships between breast cancer risk and 2-hydroxyestrone and 16a-hydroxyestrone, and the two metabolic pathways they represent, have not yet been examined in a case-control or cohort study, primarily because of the difficulty of assessing the two pathways in humans using radiolabeled tracer or gas chromatography-mass spectroscopy (GCMS) methods.
Recently, rapid and simple enzyme immunoassays (EIAs) were developed to evaluate the balance between these two metabolic pathways, and possibly to serve as a biomarker of breast cancer risk (14). The EIAs measure 2-hydroxyestrone and 166a-hydroxyestrone in unextracted urine. The initial studies that examined sensitivity, specificity, coefficients of variation, recovery, and validation, relative to a GC-MS method, have been published (14).
We hope to use these EIAs in a casecontrol study of breast cancer in Asian-American women as one of many measures of endogenous hormone levels and hormone metabolism that might explain why breast cancer incidence rates have historically been 4 to 7 times higher in the United States than in Asia (15). This large, population-based case-control study of breast cancer, conducted among women of Chinese, Japanese, and Filipino ethnicity living in San Francisco-Oakland and Los Angeles, California, and Oahu, Hawaii, was designed to take advantage of the diversity in risk and lifestyle in these ethnic populations and to elucidate the role of modifiable exposures, related to lifestyle or environment, in the etiology of this disease (15). In preparation for this effort, we examined the reproducibility and validity of a number of hormone assays that we wish to use with the plasma and urine samples collected in the study.

Collection and Distribution ofSamples
Overnight 12-hr urine samples for the hormone assays were collected from 15 volunteers working at the National Cancer Institute, Bethesda, Maryland. Midfollicular phase urine was obtained 6 to 10 days after the start of menses from five premenopausal women with regular menstrual cycles (mean age = 40 years). Midluteal phase urine was collected 4 to 6 days prior to the estimated start of the next menses from five premenopausal women with regular cycles (mean age = 39 years); subsequent follow-up confirmed the timing of the urine collection. Urine was also collected from five postmenopausal women who had experienced natural menopause (mean age = 56 years); at least three years had elapsed since their last menstrual cycle. None of the 15 women was currently taking exogenous estrogens.
The 12-hr urines were collected in halfgallon plastic jugs, containing one teaspoon of boric acid to acidify the urine and to prevent bacterial growth. The urine was kept at -40C with ice packs or refrigeration until aliquotted the next day. Each urine sample was decanted from any residue, carefully mixed, and aliquotted into 10 ml portions in 15 ml conical tubes for storage at -70°C. During March to July 1994, Bradlow's laboratory received four batches of urine aliquots, with one batch to be assayed at the beginning of each of 4 consecutive months. Each batch contained two aliquots from each of the 15 subjects. Within a batch, the 30 aliquots were randomly ordered, with a different order for each of the four batches. A year later, in March 1995, Bradlow's laboratory received a fifth batch of randomly ordered aliquots, this time with only one aliquot from each of the 15 subjects. Aliquots were shipped to the laboratory in identical, sequentially numbered vials. Laboratory personnel were told only whether an aliquot was from a premenopausal or postmenopausal woman.
Also in the spring of 1995, Adlercreutz's laboratory received a single batch of urine aliquots, containing one aliquot from each of the 15 subjects in a random order. Laboratory personnel were informed whether an aliquot was from a midfollicular premenopausal, a midluteal premenopausal, or a postmenopausal woman.

Laboratory Methods
In Bradlow's laboratory, 2-hydroxyestrone and 16a-hydroxyestrone were measured directly and concurrently in urine with anewly developed, commercially available EIA kit (Estramet 2/16, Immuna Care Corporation, Bethlehem, PA) (14). High affinity, specific murine monoclonal antibodies for each metabolite are bound to microtiter plates, and the enzyme alkaline phosphatase is linked to each metabolite. The metabolite in the sample to be assayed competes with the metabolite:alkaline phosphatase to bind to the immobilized antibody. The rate ofp-nitrophenol hydrolysis is inversely related to the concentration of the metabolite in the sample.
Initially in Bradlow's laboratory, the urine aliquots were thawed and incubated with ,B-glucuronidase/sulfatase to hydrolyze estrogen glucuronides and sulfates. Each urine aliquot was assayed in triplicate and the results averaged. Standards of 0.625 to 40.0 ng/ml (2.2-140 nmol/liter) were routinely used (14). For aliquots whose values fell off the standard curve, either twice the volume or a dilution of the urine was assayed. The between-and within-assay coefficients of variation for 2-hydroxyestrone, 16a-hydroxyestrone, and their ratio with this EIA kit have been reported to be consistently less than 9% (14).
In Adlercreutz's laboratory, GC-MS was used to characterize the estrogen profile in urine after hydrolysis of conjugates (16). Fourteen endogenous estrogensestrone, estradiol, estriol, 2-hydroxyestrone, 2-hydroxyestradiol, 2-methoxyestrone, 2-methoxyestradiol, 4-hydroxyestrone, 1 5a-hydroxyestrone, 16a-hydroxyestrone, 1 6,B-hydroxyestrone, 1 6-ketoestradiol, 16-epiestriol, and 17-epiestriol-were measured, along with four lignans and four isoflavonoids. Each urine aliquot was assayed in duplicate and the results averaged. Briefly, after protection of the carbonyl functions with O-ethylhydroxylamine, estrogen conjugates were extracted on Sep-Pak C18 cartridges (Waters Assoc., Milford, MA) and purified on the acetate form of DEAE-Sephadex. The aliquots were subsequently hydrolyzed with Helix pomatia juice and the hydrolysate purified on the acetate form of QAE-Sephadex. Recovery after hydrolysis has been estimated to be 75 to 82%, based on addition of deuterated (d5-)ethyloxime derivatives of all ketonic estrogens as internal standards immediately before this step (17). Estrogens with vicinal cis-hydroxyls and diphenolic compounds were fractionated on the borate and bicarbonate forms of QAE-Sephadex, respectively. Neutral steroids were removed by the free base form of DEAE-Sephadex, after which estrogens were separated into two groups using Lipidex 5000 in a straight phase system. Following trimethylsilyl ether derivatization, estrogens were analyzed by capillary gas chromatography with stable isotope dilution mass spectrometry. Deuterated internal standards were available for all the estrogens except 16a-hydroxyestrone and 17-epiestriol, and were used to correct for losses after the initial hydrolysis step. The limit of detection was estimated to be 0.5 to 3 nmol/liter. The coefficients of variation in premenopausal urine samples for the 10 major estrogens, including 2-hydroxyestrone and 16a-hydroxyestrone, have been reported to be 4 to 7% (16,17).

Satistical Methods
The means of the triplicate EIA readings for each aliquot were analyzed on the log scale (base 10) to reduce the dependence of the standard deviation of the response on the mean response. The transformation is also appropriate because -studies of cancer Environmental Health Perspectives * Vol 105, Supplement 3 * April 1997 association typically regress log (relative risk) on the log of assay results.
For each group of women classified by menstrual phase, we estimated components of variance among women (GYa2), among months of analysis for a given woman ((Y^), and among aliquots for a given month ((Y2).
Estimation was based on restricted maximum likelihood using the SAS procedure VARCOMP. With y#k denoting the loglo of the mean assay measurement over triplicates for woman i=1,2,3,4,5 at month j(i)= 1,2,3,4 on aliquot k(j) = 1,2, the model is

Results
Reproducibility ofEIAs for 2-and 16a-Hydroxyestrone The reproducibility of EIA measurements of urinary 2-hydroxyestrone and 16ac-hydroxyestrone over a 4-month interval is presented in Figures 1 and 2, respectively. The estrogen metabolite concentration, prior to any dilution of the urine in preparation for assay, is plotted on a logarithmic scale. Each symbol represents the mean of triplicate measurements; different symbols distinguish the five women in each menstrual group. No clear temporal trends were apparent for either metabolite over the 4-month interval. The variation in measurements taken over several months was comparable to the variation on a single day and small compared to the variation in levels between women. For both metabolites, assay variability was generally comparable in the three groups of women and at different concentrations of metabolite, with possibly somewhat greater assay variability in the postmenopausal women. For both where ai, bj(i), and Ek(ij) are independent normal variates, each with mean zero and respective variances Ga2, G2, and a2. The value of b, the month term, is taken to be different for each woman and the same for all assays in that month for that woman. The term for month is said to be nested within woman. This nesting is denoted with j(i). Similarly the effect of aliquot is nested within both month and woman, and this is denoted k(ij). The underlying mean is p and is regarded as a constant.
Under the model, two assay results from woman i done in different months will include the same parameter a.; but the terms for month and aliquot will be different and independent. Thus, the covariance between measurements arises only from the term due to the individual. The intraclass correlation coefficient is Ga2/(aG2 + Gb2 + a2), i.e., the individual variance component divided by the sum of all variance components. Coefficients of variation measure the variability associated with a laboratory and are usually estimated by repeatedly assaying aliquots from a single large pool and dividing the standard deviation of the measurements by the mean value. Here the coefficient of variation was obtained by an application of the delta method, as in Gail et al. (18). We approximated the coefficient ofvariation by 100 x 2.303 X (Gsb2 + a2 )1/2. The grand means from the 4-month reproducibility study were compared with the corresponding assay results obtained 1 year later using both a paired t test and a sign test.
Associations were examined using the Spearman rank order correlation and, after log transformation, the Pearson correlation. The Spearman rank order correlation is the usual correlation coefficient for ranked outcomes. metabolites, between-individual variability was highest for the midfollicular premenopausal women and lowest for the postmenopausal women. For women in each menstrual group, between-individual variation appeared greater for 2-hydroxyestrone than 166a-hydroxyestrone. Apart from the unusually low readings for both metabolites for one aliquot at month 3 from a midluteal premenopausal woman, a single EIA measurement, in general, reliably characterized each premenopausal woman. To see this, note that the lines connecting the midpoints of the two readings in each month for each woman are usually clearly separated (Figures 1 and 2). However, for the postmenopausal women, single EIA readings were not as reliable for characterizing an individual woman; in particular, the midpoint lines cross one another. Among the postmenopausal women, a narrower range in metabolite values and, to a lesser extent, the increased assay variability contributed to the overlapping lines and the less definitive discrimination between women.
For both 2-hydroxyestrone and 16a-hydroxyestrone, the mean of the eight EIA measurements performed on urine from each woman in the spring of 1994 was compared with the single repeat measurement made a year later (Figure 3). Readings for both estrogen metabolites increased, with the most striking effect seen in the postmenopausal women. Mean 2-hydroxyestrone readings rose in two of the midfollicular premenopausal women, in five of the midluteal premenopausal women, and in five of the postmenopausal women, while 16o-hydroxyestrone readings rose in two of the midfollicular premenopausal women, in four of the midluteal premenopausal women, and in all five of the postmenopausal women. With few exceptions, rankings among women were maintained. The mean relative increase for each estrogen metabolite was moderate but not statistically significant for the midfollicular premenopausal women, modest for the midluteal premenopausal women, and substantial and statistically significant for the postmenopausal women ( Table 1).
The reproducibility of EIA measurement of 2-hydroxyestrone and 1 6a-hydroxyestrone is summarized and quantified in Table 1. The coefficient of variation over a 4-month interval, a measure of laboratory variability, was comparable for the two estrogen metabolites and highest in the postmenopausal women (19%), the menstrual group with the lowest concentration of   both metabolites. Laboratory variation from month to month during the 4-month interval was no greater than expected, based on laboratory variation in a single day (results not shown). The intraclass correlation coefficient, the proportion of the variation in assay results attributable to differences between women, was lower in midluteal premenopausal women than in midfollicular premenopausal women, and was lowest in postmenopausal women (Table 1). In addition, it was slightly greater for 2-hydroxyestrone than 166a-hydroxyestrone. The intraclass correlation coefficient approached 99% for each metabolite in midfollicular premenopausal women (approximately 10-fold range in metabolite concentrations among the women) and dropped to approximately 80% in postmenopausal women (approximately 3-fold range in metabolite concentrations).

BiochemicalValidity ofEIAs for 2-and 16a-Hydroxyestrone
An assay may be highly reproducible but not valid. For example, the assay may detect compounds structurally similar to the analyte of interest (lack of specificity). Because of its ability to separate and identify very similar compounds (excellent specificity), GC-MS is widely accepted as a gold standard for measurement of steroid hormones. The correlations between EIA and GC-MS measurement of 2-hydroxyestrone, 16a-hydroxyestrone, and the ratio of 2-hydroxyestrone to 160x-hydroxyestrone Environmental Health Perspectives * Vol 105, Supplement 3 * April 1997  Table 2. The mean of the triplicate EIA readings obtained in the spring of 1995 from one aliquot was compared with the mean of the duplicate GC-MS readings conducted at approximately the same time on a single aliquot. Since the EIA for 2-hydroxyestrone is known to measure other C-2 hydroxylated estrogens (14), GC-MS estimates for 2-hydroxyestrone and 2-hydroxyestradiol were combined and compared with the EIA estimate for 2-hydroxyestrone. The Spearman rank order correlations for the estrogen metabolites were generally high (r . 0.9) except for 2-hydroxyestrone in postmenopausal women (r = 0.60). Pearson correlations were similar, with the lowest correlation (r = 0.60) again noted for 2-hydroxyestrone in postmenopausal women. Correlations for the 2/16 ratio were no better than those for the absolute levels of the two metabolites, and were lower for postmenopausal women than premenopausal women. Because of the limited reproducibility of the EIAs for both estrogen metabolites in postmenopausal women and the lower correlation with GC-MS estimates, regression analysis of the validity of the EIAs using GC-MS estimates as the true values was restricted to the 10 premenopausal women (Figure 4). The logarithm (base 10) of the EIA value was regressed against the logarithm of the GC-MS value to stabilize the variance. If the two measurement techniques agree perfectly, the slope of the regression line would be 1 and the intercept would be at the origin, as indicated by the dashed line in each plot. The slopes of the regressions for both metabolites were less than 1.0, indicating that both EIAs are less sensitive than GC-MS. The y-intercepts for both regressions were clearly above the origin, suggesting that both EIAs measure compounds other than the analytes of interest or detect nonspecific background. Nonetheless, despite the imperfect agreement between the absolute values generated by the two measurement techniques, the correlations between the two techniques were high (Table 2). Thus, among premenopausal women, 87 to 92% of the variance in the EIA estimates of the two metabolites and of the ratio was explained by the GC-MS measurements. Biologic Significance ofthe 2/16 Ratio Bradlow and colleagues have suggested that the 2/16 ratio, the ratio of 2-hydroxyestrone to 166a-hydroxyestrone, reliably characterizes the relative importance of the two dominant pathways for estrogen metabolism-2-hydroxylation and 16-hydroxylation ( Figure 5). To evaluate this hypothesis, we assumed that 2-hydroxyestrone and 2-methoxyestrone are the major metabolites on the 2-hydroxylation pathway and that 1 6a-hydroxyestrone, 1 7-epiestriol, and estriol are the major metabolites on the 16a-hydroxylation pathway (PK Siiteri, personal communication.) There are differences of opinion among endocrinologists about which estrogen metabolites are the dominant ones, and whether the same pattern is seen in all women. Using the GC-MS results for all 15 women, the ratio of 2-hydroxyestrone to 16a-hydroxyestrone (the 2/16 ratio) was then compared with the ratio of (2-hydroxyestrone + 2-methoxyestrone) to (16a-hydroxyestrone + 17epiestriol + estriol). The correlation was excellent (Spearman correlation coefficient = 0.90; Pearson correlation coefficient = 0.96) and is shown in the scatter plot in Figure  6A. In addition, the 2/16 ratio seemed consistently predictive; it reflected the relative importance of the competing pathways for the women in each menstrual group. In the late 1970s, Lemon and colleagues (19,20) hypothesized, on the basis of mammary carcinogenesis studies in rodents, that women with a low ratio of estriol to estrone and estradiol combined would have a high risk of breast cancer. The hypothesis was thought to explain the protective effect of full-term pregnancies (estriol is extremely high relative to the other two estrogens in the second half of pregnancy) and the higher incidence of breast cancer in Western than in Asian countries (21). However, in several casecontrol studies, urinary levels of these estrogens were determined and the calculated ratio was not convincingly related to breast cancer risk (21,22). Pike has suggested that the 2/16 ratio should be inversely related to the ratio estriol/(estrone + estradiol) and, therefore, the early epidemiologic studies refuting the estriol/ (estrone + estradiol) hypothesis from the 1970s would also tend to refute the more recent 2/16 hypothesis (MC Pike, personal communication). In Figure 6B, these two ratios are compared. The correlation was only moderate (Spearman correlation coefficient = -0.65; Pearson correlation coefficient = -0.64) and was not clearly seen in all three menstrual groups. Recently, Adlercreutz et al. (23) emphasized the evidence, derived from in vitro studies and possible mechanisms, that estrogens hydroxylated at the 2 and 4 positions may be carcinogenic. When they compared the urinary excretion of estrogen metabolites in 12 premenopausal Finnish women and 13 premenopausal recent Asian migrants to Hawaii (who would be expected to have a substantially lower risk of breast cancer than the Finnish women), excretion of 166a-hydroxylated estrogens was similar in both groups. Moreover, the significantly greater urinary excretion of total estrogens in the Finnish women was reflected in substantially elevated urinary levels of 2-and 4-hydroxylated estrogens. Total estrogen exposure has long been postulated as an underlying hormonal cause of breast cancer (1). Thus, we examined the relationship between the 2/16 ratio and total urinary estrogens ( Figure  6C). The correlation among the 15 women in our study was poor (Spearman correlation coefficient = 0.48; Pearson correlation coefficient = 0.56).

Discussion
To summarize, laboratory reproducibility of the EIAs for 2-hydroxyestrone and 16a-hydroxyestrone was good for both metabolites in premenopausal women (coefficient of variation = 8-14%) and satisfactory in postmenopausal women (coefficient of variation approximately 19%) over a 4-month interval. Laboratory reproducibility over a 12-month interval continued to be good for both metabolites in premenopausal women, but was noticeably poor in postmenopausal women, with the mean readings increasing 50 to 100%. Laboratory reproducibility was similar for the two metabolites. The sudden development of a problem at 12 months, which was not apparent during monitoring over 4 months, suggests that assay calibration, rather than analyte stability, was involved.
Wide variation in estrogen metabolite levels among the women enabled a single EIA measurement to characterize individual differences in the midfollicular phase premenopausal women (intraclass correlation coefficient = 98-99%) and in the midluteal phase premenopausal women (intraclass correlation coefficient = 85-91%). A narrower range in metabolite levels among the postmenopausal women reduced discrimination (intraclass correlation coefficient = 78-82%). For women in each menstrual group, between-individual variation, relative to laboratory variation, was greater for 2-hydroxyestrone than for 16a-hydroxyestrone. The correlation of EIA measurement with determination by GC-MS-the gold standard-was excellent (r > 0.9) for both estrogen metabolites, except for 2-hydroxyestrone in postmenopausal women. Correlations for the 2/16 ratio were no better than those for the individual metabolites, and thus also lowest for the postmenopausal women. Absolute agreement between the two measurement techniques was evaluated only in the premenopausal women. The EIAs for both metabolites were less sensitive than GC-MS, and detected nonspecific background. The nonspecific background noise tended to cancel out for the 2/16 ratio.
Thus, this EIA kit seemed to be the least useful for postmenopausal women. First, assay reproducibility, particularly long-term reproducibility, was poor in postmenopausal women, possibly due to the low concentration of both metabolites. A moderate change over time in the nonspecific background picked up by the EIAs would be the most problematic at low concentrations of metabolite. In addition, limited between-individual variation in absolute levels of both estrogen metabolites, compounded by poor reproducibility, caused the EIAs to be least able to discriminate between individuals among the postmenopausal women. Finally, the correlation between EIA and GC-MS measurements diverged the most in postmenopausal women, possibly due to the low levels of metabolite relative to background noise. In response to these concerns, new, more sensitive EIAs have been developed by TL Klug of Immunacare Corporation (Bethlehem, PA), and described by Bradlow (24). We are now evaluating the reproducibility and validity, relative to GC-MS, of these EIAs.
Certain of our conclusions were based on small numbers; in particular, our assessment of validity was based on a total of 15 women, with 5 in each menstrual group. A larger sample of women would also have strengthened our estimate of between-individual variation in levels of these two estrogen metabolites and its magnitude relative to laboratory variation.
In addition, we had no information on within-individual variation in levels of these metabolites-the degree to which different days of the menstrual cycle, menopausal status, and age may affect estrogen metabolite levels. Within-individual variability, as well as laboratory variability, must be small relative to between-individual variability for an assay to reliably characterize an individual.
Our evaluation of the EIAs for 2-hydroxyestrone and 16a-hydroxyestrone was unusually demanding. Most of the recent published articles on hormone measurement techniques have evaluated only reproducibility, not validity (25)(26)(27)(28). In addition, measurement by GC-MS may not be absolutely correct. We did not test the reproducibility of the GC-MS technique. Adlercreutz et al. (11) estimated that probably 5%, and possibly as much as 10%, of the hormone metabolites might have been lost in the GC-MS analysis because of incomplete hydrolysis prior to the introduction of deuterated internal standards.
In evaluating the validity of these EIAs, our focus was on biochemical validity, based on the published specificities of the two monoclonal antibodies. However, our ultimate interest is in biologic validitythe ability of this EIA kit, and any subsequent modifications of this kit, to predict risk of breast cancer in individual women. Even if the two antibodies are not totally specific for 2-hydroxyestrone and 16axhydroxyestrone and cross-react with other 2-substituted and 16a-substituted estrogen metabolites, respectively, the EIAs may still provide meaningful measures of the competition between these two metabolic pathways. Thus, we plan to evaluate these ELAs in a case-control study of breast cancer. By focusing attention on patterns of hormone metabolism, not just absolute hormone levels, this kit suggests new approaches for studies of endogenous hormones and breast cancer.