Estrogen metabolism in menopausal hormone users in the Women’s Health Initiative Observational Study: Does it differ between estrogen plus progestin and estrogen alone?

Abstract

The WHI found an unexpected reduced breast cancer risk in women using CEE alone. We hypothesized CEE alone induces estrogen hydroxylation along the 2-pathway rather than the competing 16-pathway, a pattern linked to reduced postmenopausal breast cancer risk. 1864 women in a WHIOS case-control study of estrogen metabolism and ovarian and endometrial cancer were studied of whom 609 were current E+P users (351 used CEE+MPA), while 272 used E alone (162 used CEE). Fifteen EM were measured, and analyses conducted for each metabolite, hydroxylation pathway (2-, 4-, or 16-pathway), and ratios of pathway concentrations using inverse probability weighted linear regression. Compared to E+P users, all EM were higher in E alone users (significant for unconjugated estrone, total/conjugated estradiol, total/unconjugated 2-methoxyestrone, 4-methoxyestrone and unconjugated estriol). The relative concentrations of 2- and 4-pathway EM did not differ between the MHT users (2-pathway EM comprised 15% and 4-pathway EM <2% of the total), but 16-pathway EM were lower in E alone users (p=0.036). Ratios of 2- and 4-pathway EM compared to 16-pathway EM were significantly higher in E alone compared to E+P users. Similar but not significant patterns were observed in CEE-alone and CEE+MPA users. Our data suggest that compared to E+P users, women using E alone have more extensive metabolism via the 2- versus the competing 16pathway. This is consistent with epidemiologic evidence of reduced postmenopausal breast cancer risk associated with this metabolic profile and may provide a clue to the breast cancer risk reduction in CEE alone users during the WHI.

Introduction

Reports from the menopausal hormone therapy (MHT) randomized trials of the WHI showed conflicting effects for women in the CEE+MPA versus CEE alone arms compared to their respective placebo groups⁽¹⁾. At termination of the CEE+MPA trial, women using CEE+MPA had a significantly elevated risk of developing breast cancer (hazards ratio (HR)=1.24, 95% confidence interval (CI)=1.01–1.53) and the longer it was used, the higher the risk⁽²⁾. Risk attenuated after discontinuation⁽³⁾ but remained elevated, even 10 years after trial cessation⁽¹⁾. Women assigned to treatment in the CEE+MPA arm experienced increased breast density⁽⁴⁾, and those that developed breast cancer were more likely to be diagnosed at a later stage⁽⁵⁾. In contrast, a non-significant breast cancer risk reduction was observed at the end of the CEE alone trial⁽⁶⁾, which became significant after trial completion, (HR=0.77, 95% CI=0.62–0.95)⁽⁷⁾, and was clearly divergent from the CEE+MPA results.

Based on a substantial body of evidence from observational studies linking breast cancer risk to use of E+P, and to a lesser degree to E alone^(8,9), the protective effect of CEE alone on breast cancer risk was unexpected. While the elevated risk for women receiving CEE+MPA is likely progestin-mediated, the underlying mechanism for a possible protective effect of CEE alone on breast cancer is not clear. We postulated that these menopausal hormone regimens may differentially influence estrogen metabolism patterns which have been implicated in breast carcinogenesis. Estrone and estradiol, the parent estrogens, are primarily hydroxylated along competing pathways at the C-2, C-4, or C-16 position of the steroid ring to produce a cascade of 2-, 4-, and 16-pathway metabolites with varying affinities for the estrogen receptor (ER)^(10,11).

Two prevailing but contradictory hypotheses regarding these metabolites and breast cancer carcinogenicity have been pursued: one, that 16-pathway metabolites are highly carcinogenic, covalently binding to the ER to induce estrogenic and mitogenic effects^(12–14) while the 2pathway metabolites weakly bind to the ER and produce anti-estrogenic effects^(15–17); the other, that catechol estrogens (2-hydroxyestrone/estradiol and in particular, 4-hydroxyestrone/estradiol) are highly carcinogenic, capable of being oxidized into mutagenic quinones that may damage DNA directly by forming quinone DNA adducts or indirectly by redox cycling⁽¹⁸⁾.

Previously, a breast cancer case-control study nested in the WHI⁽¹⁹⁾ explored whether the contradictory breast cancer risks in the trials were linked to differences in estrogen metabolism, using an enzyme-linked immunoassay (ELISA) to measure two metabolites, 2-hydroxyestrone and 16α-hydroxyestrone. Results were modest and not significant, but interpretation is hampered by assay shortcomings⁽²⁰⁾. The development of a liquid chromatography/tandem mass spectroscopy (LC-MS/MS) method to measure a panel of EM in each hydroxylation pathway, including the 4-pathway, has overcome these issues and shown promising results. A series of recent epidemiologic studies using this method found reduced postmenopausal breast cancer risk in women who preferentially metabolize estrogens along the 2- rather than the 16-hydroxylation pathway^(21–24), and in one study, an elevated risk was suggested for women with higher concentrations of 4-pathway catechol estrogens relative to the biologically inactive 4-methylated catechol estrogens⁽²²⁾. Thus, we hypothesized that altered estrogen metabolism as measured by this LC-MS/MS method, specifically higher levels of 2- relative to 16-pathway metabolites and/or higher concentrations of 4-methoxycatechol versus 4-catechol estrogens, may play a role in the observed breast cancer risk reduction among women in the CEE alone arm of the WHI.

Using data from a nested case-control study of EM in relation to ovarian and endometrial cancer risk in the WHI Observational Study (WHIOS), we explored whether levels of individual EM and/or patterns of estrogen metabolism differ between women using E alone or E+P.

Materials and Methods

Study population

Between 1993 and 1998, 93,676 postmenopausal women aged 50–79 years were recruited from 40 clinical centers in the US to participate in the WHIOS⁽²⁵⁾. Exclusions to this study included women with alcoholism, drug dependence, dementia or medical conditions with less than 3-year expected survival, or who planned to move within 3 years. At baseline visit, anthropometric measures (height, weight, waist and hip circumferences) and blood samples were collected, and self-administered questionnaires requesting information on participants’ medical, family, and reproductive history, lifestyle factors, and health behaviors including menopausal hormone therapy use and physical activity were completed. Women were instructed to bring hormone therapy medication containers to the baseline visit, where a full inventory, including information on the preparation, estrogen and progestin dose, type (tablet, capsule, patch, etc.)), and frequency of use was obtained directly from the container label. No hormonal medication restrictions were applied to this cohort.

For this analysis, all women participating in a case-control study of ovarian and endometrial cancers nested within the WHIOS were included; details have been described^(26,27). In brief, cases included all women diagnosed with ovarian or endometrial cancer between enrollment and 2012. Controls were selected from women who were cancer-free at the date of case diagnosis, and frequency matched to cases based on age at baseline (50–54, 55–59, 60–64, 65–69, 70–74, 7579), year of blood draw (1993–1996, 1997–1998), race/ethnicity (white, black, Hispanic, other/unknown), hysterectomy at baseline or during follow-up prior to the index date, and for former hormone therapy users, time since last use (≤ 1 year, > 1 year). Study participants had no history of cancer (except non-melanoma skin cancer), bilateral oophorectomy, or hysterectomy (for endometrial cancer matches only) and had ≥1.1 mL serum available. Body mass index (BMI kg/m²) at baseline was calculated from clinical data and categorized into three groups: <25.0 (normal), 25.0–29.9 (overweight), ≥30 kg/m² (obese).

Laboratory assays

Aliquoted and batched serum samples were transferred to the Laboratory of Proteomics and Analytical Technologies, Cancer Research Technology Program, Leidos Biomedical Research, Inc. (Frederick, MD) for testing. The assay quantified total (comprised of conjugated (predominantly glucuronidated and sulfated forms) plus unconjugated) concentrations of the parent estrogens (estrone and estradiol) and 13 downstream metabolites (2-hydroxyestrone, 2hydroxyestradiol, 2-methoxyestrone, 2-methoxyestradiol, 2-hydroxyestrone-3-methyl ether, 4hydroxyestrone, 4-methoxyestrone, 4-methoxyestradiol, 16α-hydroxyestrone, estriol, 16ketoestradiol, 16-epiestriol, 17-epiestriol) as well as the unconjugated concentrations of five EMs (estrogen, estradiol, estriol, 2-methoyxestrone, 2-methoxyestradiol) in serum using a stable isotope dilution LC-MS/MS assay⁽²⁸⁾. The entire panel of total EM were measured simultaneously in one run, and the panel of unconjugated EM in another. Assay reliability was monitored using masked replicates within and across batches. Coefficients of variation were <6% for all EM and intraclass correlation coefficients (ICC) ranged from 0.93–0.99^(26,27).

Statistical analyses

Inverse probability sampling weights⁽²⁹⁾ were applied to cases and non-cases, with controls weighted by the inverse sample fraction defined by the matching criteria, while cancer cases were given a weight of 1.0 because no sampling occurred, i.e., all women diagnosed with ovarian or endometrial cancer who met the inclusion criteria were selected. Because serum was collected at baseline prior to cancer diagnosis, we included both cases and controls in this cross-sectional analysis. In total, 1864 women including 383 subsequently diagnosed with ovarian, 569 with endometrial, and 16 with both cancers were studied; this analytic population represents 57,379 women when weighted back to the entire cohort. Of the 1,864 women studied, 510 cases and 473 controls were never or former menopausal hormone therapy users, while 458 cases and 423 controls used menopausal hormone therapy at blood draw.

Analyses were conducted for individual serum EM levels (pmol/L), for the sum of EM concentrations overall (total EM), the sum within each hydroxylation pathway (2-, 4-, or 16pathway EM), 2-catechols (2-hydroxyestrone, 2-hydroxyestradiol), 2-methoxycatechols (2methoxyestrone, 2-methoxyestradiol), 4-catechols (4-hydroxyestrone), 4-methoxycatechols (4methoxyestrone, 4-methoxyestradiol) and the ratio of pathways (2-/16-pathway, 2-/4-pathway, 4/16-pathway, 2-methoxycatechols/2-catechols, and 4-methoxycatechols/4-catechols). Data were log-transformed to improve normality. For each EM, geometric means and 95% confidence limits (CL) were estimated using inverse probability weighted linear regression. Standard linear regression methods were used to select the most parsimonious model, with the following variables considered as potential confounders: age at blood draw; year of blood draw; height, weight and BMI; parity; ages at menarche, first birth, and menopause; type of menopause; time since menopause; oral contraceptive use; tubal ligation; cigarette smoking; alcohol consumption; physical activity; race/ethnicity; years of education and income. Considering BMI as a continuous variable, or the inclusion of ever oral contraceptive use, height, weight, parity, ages at menarche, first birth and menopause, type of menopause, tubal ligation, alcohol consumption, physical activity, years of education and income, did not change the pattern of results. Thus, the final model included age at blood draw (50–54, 55–59, 60–64, 66–69, 70–74, 75–79 years), blood draw year (1993–1996, 1997–1998), race (Caucasian, other), smoking status (never, former, current), BMI (<25, 25–29.9, 30+ kg/m²) and time since menopause (continuous, missing set to the mean of the associated age/race/smoking status strata). Distributions of individual EM concentrations were assessed for outliers. The significance of EM differences between E+P and E alone users were assessed using contrasts.

Several sensitivity analyses were conducted. Since the intervention arms of the WHI assessed CEE alone and CEE plus MPA, we restricted analyses to women who reported baseline MHT equivalent to these formulations. Of the 609 users of E+P, 151 did not specify any formulation; 351 reported CEE + MPA at baseline [CEE (Premarin) plus MPA (medroxyprogesterone, Amen, Cycrin, Provera) or combined CEE+MPA formulations (Prempro, Premphase)]; 60 used 17βestradiol (Estrace, Climara, Civelle, Estraderm) plus MPA; 17 used esterified estrogens (Estratest, Estratab) plus MPA; 13 used estropipate (estropipate, Ogen, Ortho-Est) plus MPA; 11 used other formulations; and 6 reported formulations with E alone despite indicating use of E+P in the questionnaire. Of the 272 E alone users, 54 did not report the formulation; 162 used CEE (Premarin) alone; 12 used ethinyl estradiol (Estinyl, Estratab, Estratest); 31 used 17β-estradiol (Estrace, Estraderm, Climara); 6 used estropipate (Ortho-Est, Ogen); and 7 used CEE+MPA or other progestin formulations. In sensitivity analyses, we included only women reporting use of either CEE alone (n=162) or premarin plus MPA/combined CEE+MPA (n=351) preparations at baseline, and these models were additionally adjusted for CEE dose.

To test the robustness of results, analyses were repeated after excluding: a) women diagnosed with endometrial and ovarian cancer subsequent to blood donation and identified for the EM case-control study (n=968); and b) women with outlier EM values. For the panel of total EM (obtained simultaneously), results were unavailable for 7 women; for the unconjugated EM (also run simultaneously), 11 women did not have results (this includes the 7 women without total EM results); missing data were due to problems with specimen quality or availability. Outlier values for 32 women (5 never HRT users, 6 current E alone, 15 current E+P and 6 former users) were excluded in sensitivity analyses. Analyses were conducted using SAS version 9 software (SAS Institute, Cary, NC, USA), and all statistical tests were two-sided with 5 % type I error. No adjustment was made for multiple comparisons.

Results

The characteristics of the study population are presented (Table 1). Of the 1864 women (cases and controls) included in the study, 646 (35%), 337 (18%) and 881 (47%) were never, former and current MHT users at baseline, respectively. Among current hormone users, 609 (69%) used E+P while 272 (31%) used E alone. Participants were predominantly Caucasian (ranging from a low of 86% of never users to 97% of E alone users) and less than 10% of the study population smoked at blood draw. Women currently using MHT had higher incomes and were more likely to have used OCs than former or never users. Current E alone and E+P users were similar with regards to the other characteristics, except the latter were more likely to be younger at blood donation, college educated and have a lower BMI. Women using E alone were much younger at menopause than others, with 36% reporting cessation of menses before age 45.

Table 1.

Distribution of Women’s Health Initiative Observational Study Population by Menopausal Hormone Therapy Use and Demographic, Hormonal, Reproductive and Lifestyle Factors

		Never		Former		Current E alone		Current E+P
		N	weighted %	N	weighted %	N	weighted %	N	weighted %
		646		337		272		609
Age blood draw (mean, SD)		63.7, (7.1)		65.5, (7.3)		64.9, (7.8)		61.8, (6.4)
Controls		261	97.9	212	99.1	154	98.7	269	98.0
Ovarian Cancer Case*		143	0.9	49	0.4	80	0.9	111	0.7
Endometrial Cancer Case		234	1.4	76	0.5	37	0.4	222	1.4
Demographic Factors
Race	Caucasian	553	86.1	307	93.0	262	97.4	565	90.9
Education	High school or less	154	23.3	66	21.0	51	18.0	70	10.6
	Post high school	205	31.2	128	39.9	111	45.4	181	29.2
	College graduate	280	44.6	139	38.1	109	36.0	357	60.2
Income	<$35K	243	37.1	138	40.2	85	29.4	146	23.8
	$35K−<75K	247	39.8	118	35.6	107	40.6	259	45.2
	$75K+	106	14.2	62	18.6	60	25.6	164	25.5
Hormonal/Reproductive Factors
Recency of HT use	1−<5 years			132	33.1
	5−<10 years			71	22.6
	10+ years			134	44.3
Oral contraceptive use	ever	228	39.3	118	36.6	105	50.5	323	57.0
Hysterectomy	yes	64	12.1	58	18.6	209	89.0	8	2.2
Age at menarche	<12	157	22.3	84	25.7	60	19.6	135	23.2
	12−13	360	54.1	179	52.0	139	49.0	345	52.6
	14+	126	22.7	73	21.9	71	30.2	128	24.2
Age at menopause	<45	50	8.6	57	14.2	83	35.8	50	15.1
	45−49	114	18.8	95	23.8	74	33.5	149	29.9
	50−54	314	48.3	135	42.0	84	24.9	292	40.4
	55+	119	16.1	50	12.5	31	5.8	118	14.6
Pregnancy	never	92	15.1	42	11.7	19	7.7	71	12.8
Number births	1–2	192	29.2	120	33.6	86	35.5	244	40.5
	3–4	251	38.4	127	39.4	120	39.6	223	35.1
	5+	92	15.5	39	11.8	36	12.8	61	11.3
Age at first birth	<20	65	11.3	34	7.5	36	17.6	35	6.8
	20–29	351	50.4	208	62.1	178	60.6	387	65.3
	30+	61	9.5	23	8.5	14	3.9	50	6.7
Lifestyle Factors
Usual alcohol intake	never	77	10.3	29	10.1	30	10.0	34	6.4
	past	115	18.9	59	17.8	50	23.3	76	11.2
current drinks/week	<1	205	28.8	107	32.2	82	29.5	204	34.4
	1–<7	153	25.0	86	24.2	69	25.1	192	31.0
	7+	93	16.8	55	15.8	41	12.2	101	16.4
BMI	<25	216	43.4	142	45.0	105	35.3	342	51.1
	25–<30	190	29.7	113	31.8	107	39.2	151	27.0
	30+	238	26.4	81	23.2	59	24.9	116	21.9
Smoking	never	335	52.0	164	47.0	139	50.9	290	43.3
	past	262	38.6	145	43.3	120	43.1	292	51.3
	current	42	8.0	26	8.9	11	5.7	24	5.1
Moderate to strenuous	none	90	10.6	37	12.4	35	13.4	50	10.4
physical activity	limited	246	35.5	117	34.8	100	40.1	206	36.1
	2–3/week	112	20.1	66	20.5	54	16.5	136	20.8
	4/week	189	33.0	114	32.3	83	30.1	212	32.4

Unknown and missing values were included in the denominator for the calculation of percentages.

^*Women diagnosed with both endometrial and ovarian cancer are not included. This includes 8 never hormone users, 1 current E alone user and 7 current E+P users.

Table 2 presents absolute EM concentrations, percent pathway concentrations and ratios of pathway concentrations; significant differences are in bold font. Overall, circulating levels of EM were lowest in women who had never used MHT. While the ranges of values were comparable between never and former users, almost all EM were significantly higher in former users (p values not shown). As expected, EM in current hormone users were much higher than in never or former users (all p<0.0001, results not shown), in some instances by an order of magnitude. Among current users, all EM were higher in E alone than E+P users, and significantly so for unconjugated estrone, total and unconjugated estradiol, total and unconjugated 2-methoxyestrone, 4-methoxyestrone, and unconjugated estriol, with differences ranging from 15%−31%. EM levels were not adjusted for formulation dose.

Table 2.

Geometric Mean Concentrations of Parent Estrogens and Estrogen Metabolites According to Menopausal Hormone Therapy Use; The Women’s Health Initiative Observational Study (WHIOS)

	All Participants Menopausal Hormone Therapy Use						Women Using CEE alone or CEE+MPA Menopausal Hormone Therapy Use
	Never (646)	Former (337)	E alone (272)	E+P (609)			CEE alone (162)	CEE+MPA (351)
	Geometric means, pmol/L (95% CI)a				Δc	P-diffd	Geometric means, pmol/L (95% CI)b		Δe	P-difff
Total EM	786.0 (557.2, 1108.7)	1121.0 (781.3, 1608.4)	7457.7 (5094.9, 10905.3)	6418.9 (4469.4, 9218.8)	15%	0.135	7215.6 (4964.2, 10509.1)	7158.1 (5156.4, 9897.1)	1%	0.947
Parent Estrogens	372.4 (246.9, 6124.2)	474.9 (309.2, 728.5)	3897.5 (2480.0,6124.2)	3219.2 (2092.3, 4954.2)	19%	0.099	4080.6 (2782.2, 6002.9)	3948.1 (2846.8, 5496.3)	3%	0.808
Estrone
Total	311.7 (203.8, 476.3)	410.8 (264.0, 639.1)	3384.6 (2119.6, 5404.6)	2855.5 (1829.9, 4456.0)	17%	0.152	3442.7 (2259.7, 5271.1)	3398.8 (2352.0, 4914.8)	1%	0.918
Unconjugated	54.0 (42.4, 68.8)	53.7 (41.5,69.6)	247.2 (186.2, 327.7)	203.2 (156.0, 264.8)	20%	0.036	319.6 (223.2, 459.4)	304.1 (226.6, 407.5)	5%	0.690
Estradiol
Total	51.7 (37.5, 71.5)	54.1 (38.4, 76.3)	449.4 (311.7, 648.7)	329.6 (233.0, 466.9)	31%	0.005	573.1 (436.6, 750.0)	474.4 (393.9, 572.5)	19%	0.118
Unconjugated	13.8 (11.3, 18.1)	14.4 (11.5, 18.1)	59.6 (46.3, 76.6)	47.7 (37.8, 60.2)	22%	0.026	63.8 (46.0, 88.2)	59.4 (46.4, 76.0)		0.579
2-Pathway Estrogens	138.1 (111.3, 171.4)	204.8 (161.4, 259.8)	1025.6 (797.1, 1319.5)	906.9 (717.0, 1148.3)	10%	0.140	1006.3 (711.9, 1422.3)	980.4 (723.4, 1326.1)	3%	0.805
2-Catechols	78.1 (64.5, 94.6)	115.9 (93.0, 144.5)	598.3 (464.1, 749.2)	533.8 (430.0, 663.8)	12%	0.278	609.1 (424.1, 871.3)	585.2 (427.5, 804.3)	4%	0.734
2-Hydroxyestrone	62.3 (50.5, 76.9)	93.2 (73.5, 118.2)	476.3 (368.7, 615.2)	432.3 (341.7, 547.3)	10%	0.293	490.8 (339.0, 713.4)	472.5 (342.4, 652.0)	4%	0.752
2-Hydroxy estradiol	15.7 (13.7, 18.0)	22.3 (18.7, 26.6)	110.3 (90.5, 134.3)	99.3 (83.8, 117.6)	10%	0.247	117.9 (84.7, 164.0)	111.2 (83.7, 148.4)	6%	0.611
2-Methoxycatechols	58.6 (45.3, 75.9)	85.1 (64.6, 112.7)	407.9 (305.5, 544.6)	350.0 (265.6, 461.3)	15%	0.058	383.0 (266.1, 550.0)	368.3 (266.7, 507.8)	4%	0.713
2-Methoxyestrone
Total	37.3 (28.2, 49.3)	52.4 (39.0, 70.3)	256.2 (187.5, 349.7)	215.3 (160.1, 289.5)	17%	0.044	263.5 (179.5,387.6)	244.9 (175.4, 343.8)	7%	0.524
Unconjugated	9.0 (6.4, 12.6)	10.3 (7.2, 14.7)	71.6 (48.6, 105.4)	57.1 (39.7, 81.9)	23%	0.046	93.5 (56.4, 154.5)	86.1 (55.3, 134.3)	8%	0.604
2-Methoxy estradiol
Total	13.0 (10.1, 16.4)	19.9 (15.1, 26.3)	95.4 (71.9, 126.9)	86.8 (66.0, 114.1)	10%	0.279	66.0 (54.9, 94.6)	70.7 (55.6, 99.5)	−7%	0.532
Unconjugated	2.1 (1.6, 2.7)	2.7 (2.0, 3.6)	9.4 (7.1, 12.7)	8.9 (6.7, 11.7)	6%	0.458	11.8 (8.4, 16.4)	10.9 (8.2, 14.6)	7%	0.515
2-Hydroxy estrone-3-methyl ether	7.3 (5.9, 9.0)	10.3 (8.2, 13.0)	43.1 (33.8, 54.9)	37.9 (30.2, 47.8)	13%	0.125	46.3 (33.4, 64.1)	43.6 (32.6, 58.6)	6%	0.572
4-Pathway Estrogens	13.9 (11.0, 17.5)	23.3 (18.1, 30.0)	119.5 (91.7, 155.4)	105.1 (81.9, 134.7)	13%	0.125	99.6 (8.4, 145.5)	96.7 (69.0, 135.6)	3%	0.789
4-Catechols	7.7 (6.1, 9.6)	12.4 (9.6, 15.9)	65.6 (50.3, 85.7)	59.5 (46.4, 76.2)	10%	0.287	61.1 (41.2,90.0)	59.4 (41.9, 83.9)	3%	0.824
4-Hydroxy estrone	7.7 (6.1, 9.6)	12.4 (9.6, 15.9)	65.6 (50.3, 85.7)	59.5 (46.4, 76.2)	10%	0.287	61.1 (41.2,90.0)	59.4 (41.9, 83.9)	3%	0.824
4-Methoxycatechols	6.1 (4.8, 7.7)	10.4 (8.1, 13.5)	50.6 (38.7, 66.2)	43.0 (33.3, 55.5)	16%	0.049	37.3 (25.5, 54.6)	35.1 (24.8,49.4)	6° o	0.545
4-Methoxy estrone	4.1 (3.3, 5.1)	6.9 (5.4, 8.8)	33.7 (26.1, 43.4)	28.0 (22.0, 35.6)	18%	0.030	28.5 (19.8,40.9)	25.6 (18.5,35.2)	11%	0.313
4-Methoxy estradiol	2.0 (1.5,2.6)	3.3 (2.5, 4.4)	15.4 (11.4,20.9)	13.7 (10.2, 18.4)	12%	0.211	8.3 (5.3, 13.1)	8.7(5.7, 13.2)	−5%	0.710
16-Pathway Estrogens	244.7 (178.2, 336.3)	385.7 (275.3, 540.2)	2067.3 (1449.5, 2949.4)	1964.5 (1401.1, 2751.8)	5%	0.596	1815.3 (1163.3,2835.6)	1923.7 (1285.6,2892.9)	−6%	0.645
16-α hydroxyestrone Estriol	31.4 (24.8, 39.7)	47.6 (36.6, 61.8)	253.2 (191.5, 334.6)	233.7 (180.4, 302.8)	8%	0.395	253.9 (170.0,379.9)	252.1 (177.7,357.8)	1%	0.959
Estriol
Total	147.7 (102.6, 212.7)	241.1 (164.7, 353.2)	1320.8 (884.5, 1972.4)	1270.3 (866.1, 1865.0)	4%	0.701	1111.0 (688.8, 1790.1)	1212.0 (782.9, 1881.8)	−9%	0.513
Unconjugated	24.0 (17.4, 32.9)	35.3 (25.3, 49.3)	150.2 (106.2, 212.5)	129.4 (92.7, 180.9)	15%	0.048	168.0 (120.3,432.7)	162.1 (123.7,212.7)	4%	0.768
16-Ketoestradiol	34.8 (26.3, 46.0)	56.1 (41.4, 75.9)	308.9 (224.5, 425.4)	280.5 (214.2, 391.5)	6%	0.506	273.1 (172.4,432.7)	279.2 (184.8,424.1)	−2%	0.865
16-Epiestriol	15.1 (13.3, 17.2)	18.9 (16.0, 22.3)	85.2 (71.0, 102.3)	80.6 (68.2, 95.0)	6%	0.527	88.4 (64.9, 120.3)	90.2 (69.3, 117.9)	−2%	0.863
17-Epiestriol	12.5 (10.8, 14.4)	14.8 (12.4, 17.6)	55.4 (45.6, 67.3)	55.5 (46.6, 66.2)	0%	0.978	59.3 (44.3, 79.8)	65.6 (51.3,83.9)	−10%	0.401

^aAdjusted for age at blood draw (<55, 55–59, 60–64, 70–74, 75–79), year of blood draw (1993–1996, 1997–1998), race (Caucasian, other), smoking status (never former, current), BMI (<20, 20–24.9, 25–29.9, 30+), and years since menopause (continuous)

^bAdditionally adjusted for CEE dose.

^cPercent difference in EM between E alone and E+P users

^dP-diff refers to the p-value (Wald test) comparing the geometric mean between E alone and E+P users.

^ePercent difference in EM between CEE alone and CEE+MPA users

^fP-diff refers to the p-value (Wald test) comparing the geometric mean between CEE alone and CEE+MPA users. Missing data for the covariates was handled as a separate category.

^gAssays were conducted in two phases. For all the EM, no results were available for 7 women (never users n=1, former users n=3, E alone n=2, E+P n=1; for the unconjugated metabolites results for 11 women were also missing (former users n= 2, E alone n=2, E+P n=7). Missing data for the covariates was handled as a separate category

For the subset of women using CEE alone or CEE+MPA (Table 2), analyses were additionally adjusted for CEE dose. No significant differences were observed, although the pattern of differences persisted, with EM levels tending to be higher in users of CEE alone than CEE+MPA, particularly for 2-pathway metabolites.

To investigate patterns of estrogen metabolism, we next assessed concentrations of parent and pathway EM relative to the total (Table 3). For all women, parent estrogens (estrone and estradiol) accounted for approximately half the total EM concentration, with users of E alone having the highest relative level (in E alone and E+P users, parent estrogens comprised 54% and 52% respectively, of the total EM). Compared to never and/or former hormone users, current hormone users had a significantly lower proportion of 2-pathway metabolites (Supplemental Table 1). For 16-pathway metabolites, the relative concentrations for never, former and E+P users were similar (ranging from 32%−35%), whereas in E alone users, 16-pathway metabolites comprised a significantly lower percent of total EM (p<0.0001). Comparing E alone to E+P users, no significant differences were seen in the proportions of parent, 2-pathway, and 4pathway metabolites; however, the relative proportion of 16-pathway metabolites was 7% lower in E alone users (Table 3, p=0.036).

Table 3.

Mean Concentrations of Percent Parent Estrogens and Percent Estrogen Pathway Metabolites According to Menopausal Hormone Therapy Use; The Women’s Health Initiative Observational Study (WHIOS)

	All Participants Menopausal Hormone Therapy Usea						Women Using CEE alone or CEE+MPA Menopausal Hormone Therapy Useb
	Never (646)	Former (272)	E alone (337)	E+P (609)			CEE alone (162)	CEE+MPA (351)
	Mean (95% CI)				Δc	P-diffd	Mean (95% CI)		Δe	P-difff
Percent Parent	48.2 (44.3, 52.1)	43.5 (39.4, 47.6)	54.1 (49.6, 58.4)	51.5 (47.3, 55.6)	5%	0.074	55.0 (51.3, 58.8)	53.4 (50.4, 56.5)	3%	0.361
Percent 2-Pathway	18.1 (15.7, 20.5)	19.2 (16.7, 21.6)	14.5 (12.0, 17.1)	14.7 (12.2, 17.2)	−1%	0.714	13.5 (12.0, 14.9)	13.4 (12.1, 14.6)	1%	0.860
Percent 4-Pathway	1.9 (1.5, 2.2)	2.2 (1.9, 2.5)	1.8 (1.4, 2.1)	1.8 (1.4, 2.1)	0%	0.922	1.6 (1.4, 1.8)	1.6 (1.4, 1.7)	3%	0.636
Percent 16-Pathway	31.8 (30.3, 33.4)	35.1 (33.3, 36.9)	29.8 (27.5, 32.1)	32.1 (30.2, 34.0)	−7%	0.036	29.9 (27.2, 32.6)	31.7 (29.6, 33.7)	−6%	0.199
Percent 2-Catechols	10.3 (8.6, 12.0)	10.8 (9.1, (12.6)	8.6 (6.8, 10.4)	8.8 (7.0, 10.5)	−2%	0.592	7.9 (6.9, 8.9)	7.6 (6.9, 8.4)	1%	0.591
Percent 2-Methoxycatechols	7.8 (6.9, 8.7)	8.3 (7.4, 9.3)	5.9 (4.9, 7.0)	5.9 (5.0, 6.9)	0%	0.967	5.6 (4.6, 6.5)	5.7 (4.8, 6.6)	0%	0.707
Percent 4-Catechols	1.0 (0.8, 1.2)	1.2 (1.0, 1.3)	1.0 (0.8, 1.1)	1.0 (0.8, 1.2)	0%	0.586	0.9 (0.8, 1.0)	0.9 (0.8, 1.0)	0%	0.771
Percent 4-Methoxycatechols	0.78 (0.69, 0.87)	0.83 (0.74, 0.93)	0.59 (049, 0.70)	0.59 (0.50, 0.69)	0%	0.967	0.56 (0.46, 0.65)	0.57 (0.48, 0.66)	0%	0.707
Pathway Ratios
Ratio 2-hydroxyestrone: 16a-hydroxyestrone	1.99 (1.94,2.04)	1.97 (1.91,2.03)	1.99 (1.94,2.04)	1.89 (1.82, 1.96)	2%	0.210	1.88 (1.82, 1.95)	1.83 (1.78, 1.88)	1%	0.138
Ratio 2-/16-Pathway	0.587 (0.536, 0.639)	0.561 (0.505,0.616)	0.531 (0.468, 0.594)	0.488 (0.432, 0.547)	8%	0.048	0.512 (0.454, 0.570)	0.470 (0.425, 0.516)	8%	0.194
Ratio 4-/16-Pathway	0.058 (0.052, 0.065)	0.063 (0.056, 0.070)	0.061 (0.053, 0.069)	0.056 (0.049, 0.063)	9%	0.037	0.058 (0.051, 0.065)	0.052 (0.047, 0.057)	10%	0.087
Ratio 2-/4-Pathway	10.11 (9.80, 10.43)	8.98 (8.61,9.35)	8.75 (8.34,9.17)	8.79 (8.42,9.16)	0%	0.847	8.74 (8.22, 9.26)	8.76 (8.36, 9.17)	0%	0.924
Ratio 2-methoxycatechols: 2-catechols	0.722 (0.647,0.814)	0.674 (0.599, 0.769)	0.619 (0.545,0.715)	0.589 (0.531,0.662	−5%	0.603	0.634 (0.546, 0.756)	0.625 (0.561, 0.704)	0%	0.866
Ratio 4-methoxycatechols: 4-catechols	0.750 (0.693,0.819)	0.770 (0.696,0.861)	0.685 (0.610,0.783)	0.641 (0.584,0.710)	−7%	0.276	0.697 (0.601, 0.829)	0.662 (0.590, 0.753)	−2%	0.522

^aAdjusted for age at blood draw (<55, 55–59, 60–64, 70–74, 75–79), year of blood draw (1993–1996, 1997–1998), race (Caucasian, other), smoking status (never former, current), BMI (<20, 20–24.9, 25–29.9, 30+), and years since menopause (continuous)

^bAdditionally adjusted for CEE dose.

^cPercent difference in EM between E alone and E+P users

^dP-diff refers to the p-value (Wald test) comparing means between E alone and E+P users.

^ePercent difference in EM between CEE alone and CEE+MPA users

^fP-diff refers to the p-value (Wald test) comparing means between CEE alone and CEE+MPA users. Missing data for the covariates was handled as a separate category.

For the ratios of the EM pathways (Table 3), E alone users had approximately 10% higher concentrations of 2- and 4- relative to 16-pathway metabolites than E+P users (p=0.048 and 0.037 for 2-/16-pathway and 4-/16-pathway, respectively); however, the ratios of the metabolites 2-hydroxyestrone/16α-hydroxyestrone were comparable. The relative concentration of 4catechols compared to 4-methoxycatechols (and 2-catechols relative to 2-methoxycatechols) did not differ between the MHT groups.

In analyses restricted to CEE alone and CEE+MPA users, no significant differences were observed for the proportion of parent, 2-, 4-, and 16-pathways relative to total EM, but the patterns observed in the full study persisted (Table 3). The relative concentration of 16-pathway EM continued to be lower in users of CEE alone compared to CEE + MPA, and CEE alone users had higher levels of 2- compared to 16-hydroxylation pathway EM (p>0.05).

Several sensitivity analyses were conducted. Restricting analyses to the 896 women without ovarian or endometrial cancer did not alter the pattern of results, although differences between E alone and E+P users were significant only for the parent estrogens, the 2- and 4methoxycatechols, the relative proportion of 16-pathway metabolites, and the ratio of 2-:16pathway metabolites (Supplemental Table 2). Among the non-cases, analyses limited to CEE or CEE+MPA users showed similar patterns as the full study, albeit with no significant findings.

Excluding outliers did not change the interpretation of findings (results not shown).

Discussion

Our data suggest that compared to women using E+P, E alone users may be less likely to metabolize estrogens along the 16-pathway, and thus have more extensive metabolism along the 2- pathway relative to the competing 16-pathway. A similar pattern was seen with E alone users having higher 4-/16-pathway metabolism than E+P users. No significant differences in estrogen metabolism were seen in analyses restricted to women using CEE alone or CEE+MPA; however, as was the case for all E alone users, lower 16-pathway hydroxylation was suggested for users of CEE alone. EM concentrations were highest in women using E alone, which may in part, reflect different estrogen formulations and dosages between the MHT users; however, this could not be assessed since this information was unavailable for approximately 25% of participants. In the subset with MHT formulation and dose, analyses adjusted for dose continued to show all EM, including 4-methoxyestrone, to be higher in CEE alone users.

Our findings are in accordance with some, but not all, results from the earlier breast cancer case-control study nested within the WHI⁽¹⁹⁾ which used an ELISA to measure 2- and 16αhydroxyestrone. Similar to our findings, after one year of treatment, CEE+MPA users in that study experienced a larger increase in 16α-hydroxyestrone than users of CEE alone but increases in 2-hydroxyestrone were comparable for both groups. In women randomized to treatment, breast cancer risk was non-significantly reduced in CEE alone users who experienced increases in both 2-hydroxyestrone and in the ratio of 2-/16α-hydroxyestrone. In contrast, in serum collected prior to study randomization, elevated breast cancer risk was associated with higher levels of both 2-hydroxyestrone and the ratio of 2-/16α-hydroxyestrone. Unlike the LC-MS/MS method, the ELISA was designed primarily to evaluate the relative dominance of competing estrogen metabolic pathways, using the ratio of 2-hydroxyestrone and 16α-hydroxyestrone as a proxy for those pathways. Yet comparability of the LC-MS/MS and ELISA methods is in doubt, particularly for postmenopausal women, since large differences in the absolute values of the metabolites have been shown⁽³⁰⁾ and the correlation between the calculated ratios of 2-/16αhydroxyestrone by each method is weak⁽³⁰⁾. Additionally, our LC-MS/MS data suggest that the ELISA ratio of 2- and 16α-hydroxyestrone alone does not adequately capture the dominance of one pathway over the other, since the concentration of 2-hydroxyestrone was twice as high as 16α-hydroxyestrone, yet the total concentration of all 2-pathway components was approximately half that of the 16-pathway metabolites. Taken together, these observations complicate the interpretation of findings from earlier studies using the ELISA and may in part explain the inconsistent and largely inconclusive epidemiologic evidence linking these two metabolites to breast cancer risk^(20,24,31).

A smaller body of recent evidence from studies using this LC-MS/MS method⁽²⁴⁾ demonstrates breast cancer is reduced in postmenopausal women with more extensive metabolism along the 2pathway rather than via the competing 16-pathway. Thus, the shift in estrogen metabolism away from the 16-pathway, and the resulting increase in 2- relative to 16-pathway metabolism we found in E alone users may provide a partial explanation for the breast cancer risk reduction observed in these women during the WHI. While we did not observe differences between MHT groups in the relative concentration of 4-methoxycatechols compared to 4-catechols, 4methoxyestrone was significantly higher in E alone users. Methylation of catechols prevents further metabolism of catechol estrogens to catechol estrogen quinones, thereby deactivating the pathway that produces reactive and potential mutagenic metabolites.

Several caveats must be borne in mind when interpreting our results. Perhaps most concerning, our LC-MS/MS assay did not quantitate most of the estrogens in the MHT formulations, and the biological relevance of these steroids has not been well studied in humans. CEE is a complex extract of pregnant mares’ urine containing several estrogens, progestins, and androgens. At least 10 different estrogens have been identified including the ring B saturated estrogens including estrone (comprising 50% of the CEE dose), 17β-estradiol (<1%), and 17α-estradiol (4.5%), and the ring B unsaturated estrogens: equilin (22%), equilenin (2%), 17α-dihydroequilin (14%), 17β-dihydroequilin (1.7%), 17α-dihydroequilenin (<1%), 17β-dihydroequilenin (<1%), and delta-8-estrone (3.5%)⁽³²⁾. While our assay may have captured estrone, other known estrogens in the formulation were not measured, and the activity of these unmeasured estrogenic (as well as androgenic and progestogenic) components in CEE is not well understood. CEE may contain upwards of 200 distinct compounds.

Since the WHIOS did not restrict menopausal hormone therapy use at baseline and women reported a variety of estrogen and progestin formulations, concerns regarding the comparability of the WHI trial and WHIOS populations are reasonable. We note that levels of the parent estrogens in our study were similar to those in WHI participants in the CEE alone and CEE+MPA arms following the first year of treatment⁽³³⁾, and as discussed, our finding that E alone users were less likely to metabolize estrogens along the 16-pathway had been observed in CEE alone vs CEE+MPA participants in the WHI⁽¹⁹⁾. In addition, the WHI trial breast cancer findings of excess risk for E+P users and a reduced risk for E alone users, have been corroborated in the WHIOS cohort^(34–36). Finally, our decision to include approximately 50% of the study population who were healthy at the time of blood draw for the WHIOS but later diagnosed with endometrial or ovarian cancer is not likely to have influenced results for this study, since these women accounted for only 2% of the study population in the weighted analysis. Sensitivity analyses excluding these women did not alter our interpretation, although results were no longer significant.

In conclusion, our data suggest that estrogen metabolism may differ for E alone compared to E+P users, with E alone inducing estrogen metabolism away from the 16α-pathway and towards the 2-pathway, a pattern shown to reduce breast cancer risk in postmenopausal women. However, the magnitude of observed differences between these MHT groups is small, and the chemical complexity of their formulations may confound the interpretation of findings regarding estrogen metabolism. Whether this pattern of estrogen metabolism persists following cessation of hormone therapy is not known.

Novelty and Impact:

The WHI found an adverse effect on breast cancer risk for CEE+MPA, but an unexpected reduced risk in the CEE alone arm; the latter runs counter to the known carcinogenic effect of estradiol. Here the authors found CEE alone may induce estrogen metabolism favoring the 2-hydroxylation pathway over the 16-pathway, using EM data from case-control studies of ovarian and endometrial cancer in the companion WHIOS. This pattern of metabolism has been linked to reduced postmenopausal breast cancer risk.

Abbreviations:

MHT

menopausal hormone therapy

CEE

conjugated equine estrogens

CEE+MPA

conjugated equine estrogens plus medroxyprogesterone acetate

WHI

Women’s Health Initiative Clinical Trial

WHIOS

Women’s Health Initiative Observational Study

EM

estrogen metabolites

E alone

estrogen alone

E+P

estrogen plus progestin