Circulation 2003 Rexrode 1688 93

Sex Hormone Levels and Risk of Cardiovascular Events inPostmenopausal Women

Kathryn M. Rexrode, MD, MPH; JoAnn E. Manson, MD, DrPH; I-Min Lee, MD, SCD;Paul M Ridker, MD, MPH; Patrick M. Sluss, MD; Nancy R. Cook, ScD; Julie E. Buring, ScD

BackgroundDespite diffuse effects of sex hormones on the cardiovascular system, few prospective studies haveexamined the relationship of plasma androgens and estrogens with risk of cardiovascular disease (CVD) inpostmenopausal women.

Methods and ResultsA nested case-control study was performed among women in the Womens Health Study. Twohundred women who developed CVD were matched 1:1 by age, smoking, and postmenopausal hormone therapy (HT)to controls who remained free of CVD. We measured testosterone, estradiol, and sex hormone binding globulin (SHBG)levels and calculated free androgen index (FAI), free estradiol index, and the FAI/free estradiol index ratio. Results werestratified by HT use. Among HT nonusers, cases had significantly higher androgen profiles (higher median FAI andlower SHBG levels) than controls. After adjustment for age, smoking, use of aspirin, vitamin E, and alcohol, familyhistory of myocardial infarction, and physical activity, nonusers in the lowest SHBG quartile had an OR of 2.25 (95%CI, 1.03 to 4.91) for CVD, and there were significant trends across FAI quartiles (P for trend0.03). Additionaladjustment for body mass index, hypertension, diabetes, and elevated cholesterol eliminated associations with SHBGand FAI. Among women using HT, no significant differences in hormones or SHBG were observed among women whodeveloped CVD and controls.

ConclusionsAmong HT nonusers, lower SHBG and higher FAI levels were noted among postmenopausal women whodeveloped CVD events, but this was not independent of body mass index and other cardiovascular risk factors. Estradiollevels were not associated with risk of CVD in HT users or nonusers. (Circulation. 2003;108:1688-1693.)

Key Words: hormones women cardiovascular diseases risk factors

Several lines of indirect evidence suggest that sex hormonelevels may be associated with risk of cardiovasculardisease (CVD) in women. In biologic studies, estrogen hasdiffuse effects on the cardiovascular system, including favor-able effects on lipid profiles and fibrinolytic proteins,1 as wellas adverse effects on inflammatory and thrombotic markers.2

Although exogenous hormones have recently been docu-mented to increase risk of CVD,3 the role of endogenousestrogen levels as a risk factor for CVD in postmenopausalwomen has not been well studied. Whether estradiol levelsinfluence risk of CVD among women undergoing hormonetherapy (HT) has not been determined.

High androgen levels may increase cardiovascular risk inwomen through adverse effects on lipids, blood pressure, andglucose metabolism.4,5 In one cross-sectional study, testoster-one levels correlated with the degree of coronary atheroscle-rosis measured on angiography.6 Although endogenous sexhormone levels have been associated with risk of osteoporo-

sis7 and breast cancer,8 only 2 small prospective studies911

have investigated the relationship of sex hormone levels andrisk of CVD in postmenopausal women. The higher andro-gen/estrogen ratio in postmenopausal than premenopausalwomen has been posited as 1 mechanism for increasedcardiovascular risk after menopause,4 yet no prospectivestudies have looked at this as a predictor of cardiovascularevents in postmenopausal women.

The Womens Health Study provided a large cohort inwhich to analyze the relationship of plasma sex hormonelevels (testosterone and estradiol) as well as sex hormonebinding globulin (SHBG) and risk of cardiovascular eventsamong postmenopausal women. Our primary hypothesis atthe time the study was designed was to test whether lowerestrogen levels (low total estradiol and low free estrogenindex [FEI]), lower SHBG levels, and higher androgen levels(high total testosterone, high free androgen index [FAI], andhigh FAI/FEI ratio) predict risk of CVD in postmenopausalwomen.

Received September 23, 2002; de novo received May 20, 2003; revision received July 24, 2003; accepted July 26, 2003.From the Department of Medicine, Division of Preventive Medicine (K.M.R., J.E.M., I.-M.L., P.M.R., N.R.C., J.E.B.), Brigham and Womens Hospital

and Harvard Medical School; Department of Epidemiology (J.E.M., I.-M.L., J.E.B.), Harvard School of Public Health; Reproductive Endocrine UnitAssay Laboratory (P.M.S.), Massachusetts General Hospital; and Department of Ambulatory Care and Prevention (J.E.M., J.E.B.), Harvard MedicalSchool, Boston, Mass.

Correspondence to Dr Kathryn M. Rexrode, Division of Preventive Medicine, Department of Medicine, Brigham and Womens Hospital, 900Commonwealth Ave E, Boston, MA 02215. E-mail [email protected]

2003 American Heart Association, Inc.

Circulation is available at http://www.circulationaha.org DOI: 10.1161/01.CIR.0000091114.36254.F3

1688 by guest on July 16, 2015http://circ.ahajournals.org/Downloaded from
http://circ.ahajournals.org/

MethodsThe Womens Health Study (WHS) is an ongoing, randomized,double-blind, placebo-controlled study of low-dose aspirin andvitamin E in the prevention of cardiovascular events and canceramong 39 786 female health professionals age 45 years or older atbaseline.12 Women complete yearly questionnaires on demographic,behavioral, and lifestyle factors as well as the occurrence of anymedical illnesses. Data on traditional cardiovascular risk factorswere gathered at baseline. Blood samples were collected in citratetubes from 28 345 women (71%) before randomization (in years1993 to 1996) and were stored in liquid nitrogen freezers at 170Cuntil the time of analysis.

Annually, WHS participants report the occurrence of any cardio-vascular outcome and consent to release of relevant medical records,which are reviewed by an end points committee of physicians.Myocardial infarction (MI) is the presence of characteristic symp-toms and either elevated cardiac enzymes or ECG changes. Coronaryrevascularization (percutaneous transluminal coronary angioplastyor coronary artery bypass grafting) is confirmed by documentation inthe medical record. Deaths attributable to coronary disease areconfirmed on the basis of autopsy reports, symptoms, circumstancesof death, and prior history of coronary disease. Stroke is confirmedby documentation of a typical neurologic deficit of sudden or rapidonset that persists for 24 hours or until death. A composite endpoint of CVD, defined as the first occurrence of nonfatal MI,coronary revascularization, nonfatal stroke, coronary disease, orstroke death, was used.

All women in the WHS were free from known prior coronary heartdisease, stroke, or cancer (except for nonmelanoma skin cancer) atstudy entry. Two hundred cases occurred among WHS participantswho (1) were postmenopausal, (2) had citrated plasma samplesavailable, and (3) suffered a confirmed CVD event. For each case, acontrol subject was selected from the WHS who met the samecriteria but remained free from CVD during follow-up and wasmatched to the index case by age, smoking status, and use of HT(never, past, and current).

All assays were conducted without knowledge of the case/controlstatus. Matched case-control pairs were handled identically andassayed in the same analytical run. Twenty quality-control speci-mens (n20) were interspersed among the study samples. Over a3-year period, intraclass correlation coefficients were 0.73 for freeestradiol, 0.92 for SHBG, and 0.88 for testosterone in postmeno-pausal female nurses.13

Estradiol levels were measured at Quest Diagnostic in 2000 byradioimmunoassay preceded by extraction and purification by Celitecolumn chromatography.14 The lower limit of detection for the assaywas 5 pg/mL; the coefficient of variation was 9.5%. Estradiol levelsfrom citrated plasma were slightly lower than from EDTA, with acorrelation of 0.89. The linear regression formula was (plasma citrateestradiol level)0.74 (EDTA plasma estradiol)8.6.

Testosterone and SHBG levels were assayed at the MassachusettsGeneral Hospital Reproductive Endocrine Laboratory in 2000. Totaltestosterone was measured using a solid-phase radioimmunoassay(Diagnostic Products Corporation). The lower limit of sensitivitywas 4.0 ng/dL. SHBG was measured using a fully automated system(Immulite; Diagnostic Products Corporation), which used a solid-phase, 2-site immunometric assay. Coefficients of variation were4.5% for SHBG and 4.8% for total testosterone.

To estimate free (nonprotein-bound) testosterone, we calculatedthe free androgen index (FAI), the molar ratio of total testosterone/SHBG, which is highly correlated with free testosterone15 (to converttestosterone to nanomolar, the nanogram per deciliter value ismultiplied by 0.0347). Similarly, to estimate the free estradiolconcentration, the free estradiol index (FEI), the molar ratio ofestradiol/SHBG was calculated (to convert estradiol to picomolar,the picogram per milliliter value is multiplied by 3.67).

Because of skewed distributions, hormone medians were com-pared using the Wilcoxon rank-sum test. Quartiles were createdbased on the distribution of values in the control population. Becauseof the a priori hypothesis that low estrogen levels and high androgenlevels were associated with increased risk of CVD, referent catego-

ries were defined as high estrogen quartiles (high estradiol and FEI)and low androgen quartiles (high SHBG and low testosterone, FAI,and FAI/FEI ratio). Because the distribution of hormonal valuesdiffered significantly for present HT users and past/never users,analyses were stratified by HT status and separate quartiles werecreated. Values for past users and never users were similar; thus,these 2 groups were combined into nonusers.

Logistic regression was performed to adjust for matching factors,traditional risk factors, and potential confounders. Data on age, bodymass index (BMI), physical activity (rarely or never, 1 to 3 times permonth, 1 time per week), smoking (never, past, current), alcoholconsumption (rarely or never, 1 to 3 times per month, 1 time perweek), HT history, parental history of MI before age 60 years, aswell as medical history of diabetes, elevated cholesterol, andhypertension were obtained from the WHS questionnaires. Modelswere also adjusted for randomized aspirin and vitamin E treatmentassignment. Test for trend across quartiles was performed usingordinal variables. Conditional and unconditional logistic regressionanalyses yielded nearly identical results; thus, the unconditionallogistic regression results are presented.

ResultsOf the 200 women who developed cardiovascular eventsduring a median follow-up of 2.9 years, there were 88nonfatal MIs, 23 coronary revascularizations, 62 nonfatalischemic strokes, and 27 cardiovascular deaths (5 fatalstrokes and 22 coronary heart disease deaths). The averageage was 63.1 years, and 42.5% were current HT users, 24.5%past users, and 33.0% never users. In the total population,median levels of SHBG, testosterone, FAI, estradiol, FEI,FAI/FEI, and testosterone/estradiol did not significantly dif-fer among women with cardiovascular events and those whoremained free of cardiovascular events (data not shown).

Because of substantial effect modification by HT, allanalyses were stratified by HT status. Among nonusers,women who developed cardiovascular events had higherBMI, tended to be less physically active, and were morelikely to have a history of diabetes, hypertension, or elevatedcholesterol than women without cardiovascular events (Table1). Among current HT users, cases were more likely to havea history of hypertension. As expected, hormone levelsdiffered according to HT use (Table 2). Although totaltestosterone levels and FEI were similar among users andnonusers, SHBG and estradiol were higher and FAI levelsand FAI/FEI were lower. Therefore, separate quartiles werecreated for HT users and nonusers.

Among women not using hormones, median SHBG levelswere significantly lower (P0.03) and FAI was significantlyhigher (P0.008) among the 115 women who developed CVDevents than controls (Table 3). No other hormones differedsignificantly. Among current HT users, there were no differ-ences between the 85 women with subsequent cardiovascularevents and those women who remained free from CVD.

Logistic regression was performed to control for matchingfactors, as well as randomized aspirin and vitamin E assignment(Table 4). Among HT nonusers, higher androgens were associ-ated with significant trends for CVD, with increasing riskassociated with lower SHBG levels (P0.03) and higher FAI(P0.02). Women in the lowest SHBG quartile had an OR of2.45 (1.14 to 5.23), whereas those in the highest FAI quartile hadan OR of 1.94 (95% CI, 0.91 to 4.14) compared with women in

Rexrode et al Sex Hormone Levels and Risk of CVD in Women 1689

by guest on July 16, 2015http://circ.ahajournals.org/Downloaded from

the lowest quartile. Low estradiol levels were not associated withincreased risk (OR, 1.09; 95% CI, 0.50 to 2.35).

After adjustment for parental history of MI, alcohol use,and physical activity, significant associations remained forwomen in the lowest SHBG quartile, with an OR of 2.25(95% CI, 1.03 to 4.91) with a significant trend acrossquartiles (P for trend0.05). Trends were also significantacross FAI quartiles (P for trend0.03). Additional adjust-ment for BMI, which is strongly and inversely associatedwith SHBG, as well as history of elevated cholesterol,diabetes, and hypertension eliminated independent associa-

tions for both SHBG and FAI. Trends for FAI/FEI were ofborderline significance, with highest risk observed amongthose in the highest quartile.

We hypothesized that hormone levels attained among HTusers might be also associated with cardiovascular risk.However, among current HT users, no significant trends orassociations were seen.

DiscussionAmong postmenopausal women not taking HT, we found thatwomen with low SHBG or high FAI were at increased risk of

TABLE 1. Baseline Characteristics of Women With and Without Cardiovascular Disease Events DuringFollow-Up According to HT Use

Characteristic

HT Nonusers Current HT Users

Women WithCardiovascular

Events (n115)

Women Free ofCardiovascular

Events (n115)P

Value*

Women WithCardiovascularEvents (n85)

Women Free ofCardiovascularEvents (n85)

PValue

Mean age, y 64.6 64.7 60.9 61.0

Smoking, %

Never 40.0 40.0 43.5 43.5

Past 39.1 39.1 32.9 32.9

Current 20.9 20.9 23.5 23.5

Mean BMI, kg/m2 28.2 26.3 0.007 26.1 24.9 0.08

Mean age at menopause, y 47.5 46.9 0.48 46.9 46.2 0.46

Physical activity, %

Rarely/never 61.7 42.6 0.01 36.5 30.6 0.08

1 to 3 times per month 14.8 18.3 28.2 17.7

1 time per week 23.5 39.1 35.3 51.8

Alcohol consumption, %

Rarely/never 49.6 57.4 0.32 45.9 49.4 0.86

1 to 3 times per month 16.5 10.4 15.3 12.9

1 time per week 33.9 32.2 38.8 37.7

History of oophorectomy, % 18.3 13.9 0.37 37.7 38.8 0.88

Parental history of MI, % 16.4 9.9 0.16 15.9 11.1 0.38

History of diabetes, % 17.4 3.5 0.001 11.8 4.7 0.09

History of hypertension, % 58.8 35.4 0.001 59.0 40.0 0.01

History of elevated cholesterol, % 51.3 30.4 0.001 40.0 36.5 0.64

*P value from 2 and t tests.Matched in case-control selection process.History of MI in either or both parents before age 60 years.History of BP 140/90.History or treatment of high cholesterol.

TABLE 2. Comparison of Median Hormone Levels by HT Use

Hormone Median (Interquartile Range)

Total Population HT Nonusers Current HT Users P Value*

SHBG, nmol/L 61.5 (34.8106) 40.8 (29.164.2) 117.5 (66172) 0.0001

Testosterone, ng/dL 18 (1428) 19 (1428) 18 (1326) 0.58

FAI, nmol/L 0.010 (0.0050.021) 0.017 (0.0090.03) 0.006 (0.0030.01) 0.0001

Estradiol, pg/mL 14 (1025) 11 (915) 26 (1645) 0.0001

FEI, nmol/L 0.0010 (0.00050.0017) 0.0010 (0.00050.0017) 0.0009 (0.00060.0017) 0.24

FAI/FEI 12.31 (6.5419.8) 17.4 (11.826.0) 6.50 (3.4711.7) 0.0001

*Comparison of nonusers to current HT users.

1690 Circulation October 7, 2003


CVD events. In age-adjusted analyses, low SHBG levelsconferred a 2-fold increased risk of cardiovascular events;however, this was mediated primarily by BMI and othercardiovascular risk factors and was eliminated after multivar-iate adjustment. By contrast, for current HT users, there wereno clear associations between estrogen, androgen, or SHBGlevels and risk of cardiovascular events.

Low estrogen levels had been proposed as an explanationfor the dramatic rise in cardiovascular risk associated withmenopause. Recent randomized trial evidence3 demonstratedincreased risk of CVD with HT, raising concern that higherestrogen levels might increase risk. The relationship ofendogenous estrogen levels to risk of CVD in postmeno-pausal women has not been frequently examined. We alsofound no relationship between estradiol levels and risk ofCVD in HT users or nonusers.

Higher testosterone levels in women have been associatedwith dyslipidemia5 as well as type 2 diabetes mellitus16 andhypertension.17 During menopause, a greater decline in es-trogen levels than androgen levels results in a higher andro-gen/estrogen ratio in postmenopausal than premenopausalwomen. There were weak trends toward increased risk ofCVD among women with higher androgen/estrogen (FAI/FEI) ratios.

Prior direct data on endogenous hormone levels and risk ofCVD are limited. In a case-control study, women withpolycystic ovary syndrome, a condition associated withelevated androgen levels, had more extensive atherosclerosison angiography than women with normal ovaries.18 In an-other angiographic case-control study, free testosterone was astronger predictor of the degree of atherosclerosis than wastotal cholesterol. Estradiol was inversely but nonsignificantlyrelated to degree of atherosclerosis.6 In contrast, atheroscle-rosis as measured by carotid intimal-medial thickness wasinversely associated with testosterone in 2 cross-sectionalstudies.19,20

Before our study, only 2 small prospective studies hadexamined the relationship between endogenous hormonesand CVD, and neither found significant relationships betweenendogenous hormones and CVD. In the Rancho BernardoStudy of 651 women, age-adjusted mean levels of bioavail-able testosterone and estradiol9 did not predict CVD deaths

during 19 years of follow-up. In a small prospective study,both estradiol and testosterone levels were nonsignificantlylower among 40 postmenopausal diabetic women who died ofcoronary heart disease than diabetic women who remainedfree of coronary heart disease.11 Our results support a lack ofassociation between estradiol and risk of cardiovascularevents.

Data for the relationship between SBHG and CVD havebeen mixed. Low SHBG levels, which are sometimes con-sidered an androgenic marker in women, have been associ-ated with low HDL cholesterol11 as well as increased risk ofdiabetes.21 In case-control studies, low plasma levels ofSHBG were associated with higher likelihood of atheroscle-rosis on angiography22 and carotid intimal-medial thick-ness20; however, another angiographic case-control studyfailed to find an association.23 Two prospective studies haveexamined the relationships between SHBG and cardiovascu-lar outcomes. In the Gothenburg Study, low SHBG concen-trations were associated with increased overall mortality andMI during 12 years of follow-up but were not adjusted forBMI.24 In the Rancho Bernardo Study, which did adjust forBMI, SHBG levels were not associated with cardiovascularmortality.10 In our study, low SHBG was associated withsignificantly higher risk of CVD events among women whodid not use HT; however, this relationship was not indepen-dent of BMI.

This study has several strengths and limitations. To ourknowledge, it is the largest prospective study of endogenoushormone levels and risk of CVD in postmenopausal women.In addition, because of the differences in hormone levelsbetween HT users and nonusers, analyses were stratified byHT use; however, we had limited power to find an associationin these separate HT strata. Although only 1 measure ofhormones was available, levels have been shown to be quitestable in postmenopausal women.13 Additionally, we mea-sured only total hormone levels, but the calculated free indexhas been shown to be a good estimate of free levels.

The role of exogenous hormones on cardiovascular risk hasbeen recently reported, but whether endogenous hormonesalso have an effect on this risk has not been well tested. Thepresent study found that among women who were not usingHT, low SHBG levels were a marker of cardiovascular risk,

TABLE 3. Median Baseline Hormone Levels Among Women With and Without Cardiovascular Events, Accordingto HT Use

Hormone Median (Interquartile Range)

HT Nonusers Current HT Users

Women WithCardiovascular

Events (n115)

Women Free ofCardiovascular

Events (n115)P

Value

Women WithCardiovascularEvents (n85)

Women Free ofCardiovascularEvents (n85)

PValue

SHBG 36.5 (25.758.8) 48.5 (32.471.1) 0.03 106 (66.5175) 126 (66171) 0.28

Testosterone 19 (1530) 18 (1427) 0.42 19 (1328) 18 (1423) 0.40

FAI 0.02 (0.010.03) 0.01 (0.0080.03) 0.008 0.006 (0.0030.01) 0.005 (0.0040.010) 0.88

Estradiol 11 (815) 11 (914) 0.44 26 (1743) 25 (1546) 0.81

FEI 0.001 (0.0060.003) 0.0009 (0.00050.002) 0.07 0.0009 (0.00070.002) 0.0008 (0.00060.002) 0.44

FAI/FEI 17.8 (12.428.2) 15.7 (11.121.8) 0.32 7.57 (3.4712.2) 5.61 (3.3611.5) 0.09



but this risk was not independent of BMI, hypertension, anddiabetes. There were weak trends toward increased riskamong women with higher FAI/FEI ratios. We did not findany association between estradiol levels and risk of CVDeither among HT users or nonusers. Larger studies of therelationship between hormone levels and risk of CVD inwomen who are not using exogenous hormones are necessaryto determine whether androgenicity is an independent cardio-vascular risk factor in postmenopausal women.

AcknowledgmentsThis study was supported by grants CA47988 and HL43851 from theNational Institutes of Health and by the Doris Duke CharitableFoundation. The authors would like to acknowledge the crucialcontributions of the entire staff of the WHS, under the leadership of

David Gordon, Susan Burt, Mary Breen, Marilyn Chown, RoseUngarelli, Lisa Fields-Johnson, Georgina Friedenberg, Jean Mac-Fadyen, Geneva McNair, David Potter, Claire Ridge, and HarrietSamuelson, as well as Natasha Gomelskaya for her assistance withdata analysis. Finally, the authors are deeply indebted to thededicated and committed participants of the WHS.

References1. Mendelsohn ME, Karas RH. The protective effects of estrogen on the

cardiovascular system. N Engl J Med. 1999;340:18011811.2. Cushman M, Legault C, Barrett-Connor E, et al. Effect of postmeno-

pausal hormones on inflammation-sensitive proteins: the PostmenopausalEstrogen/Progestin Interventions (PEPI) Study. Circulation. 1999;100:717722.

3. Risks and benefits of estrogen plus progestin in healthy postmenopausalwomen: principal results from the Womens Health Initiative randomizedcontrolled trial. JAMA. 2002;288:321333.

TABLE 4. Multivariate OR of Cardiovascular Events by Quartile of Hormone Level in HT Nonusers and Current Users

Quartiles(Q) ofHormone

HT Nonusers Quartiles(Q) of

Hormone

Current HT Users

Model 1* Model 2 Model 3 Model 1* Model 2 Model 3

SHBG, nmol/L

Q1 (32.4) 2.45 (1.145.23) 2.25 (1.034.91) 1.18 (0.462.99) Q1 (64.5) 0.89 (0.372.11) 0.99 (0.402.44) 0.66 (0.231.90)

Q2 (32.448.5) 1.17 (0.522.64) 1.12 (0.482.61) 0.73 (0.271.96) Q2 (64.5120) 1.23 (0.522.91) 1.40 (0.563.49) 0.99 (0.362.73)

Q3 (48.569.5) 1.42 (0.643.17) 1.27 (0.562.91) 1.42 (0.573.54) Q3 (120171) 0.71 (0.301.72) 0.73 (0.291.80) 0.56 (0.211.49)

Q4 (69.5) 1.00 referent 1.00 1.00 Q4 (171) 1.00 1.00 1.00

P for trend 0.03 0.05 0.94 0.90 0.70 0.69

Testosterone, ng/dL

Q1 (14) 1.00 referent 1.00 1.00 Q1 (13) 1.00 1.00 1.00

Q2 (1418) 1.33 (0.622.83) 1.22 (0.562.66) 1.40 (0.583.36) Q2 (1318) 0.41 (0.161.07) 0.38 (0.141.01) 0.23 (0.080.71)

Q3 (1826) 1.22 (0.572.63) 1.15 (0.522.52) 0.99 (0.402.42) Q3 (1824) 0.65 (0.271.55) 0.61 (0.251.49) 0.48 (0.181.28)

Q4 (26) 1.58 (0.753.34) 1.39 (0.643.00) 1.16 (0.482.78) Q4 (24) 1.24 (0.552.80) 1.05 (0.442.47) 0.74 (0.301.87)

P for trend 0.28 0.46 0.92 0.44 0.73 0.87

FAI

Q1 (0.008) 1.00 referent 1.00 1.00 Q1 (0.003) 1.00 referent 1.00 1.00

Q2 (0.0080.013) 0.68 (0.291.59) 0.59 (0.251.44) 0.58 (0.221.54) Q2 (0.0030.005) 0.69 (0.291.65) 0.68 (0.281.69) 0.65 (0.251.71)

Q3 (0.010.026) 1.82 (0.863.82) 1.68 (0.783.63) 1.34 (0.533.41) Q3 (0.0050.009) 0.77 (0.321.88) 0.75 (0.291.93) 0.66 (0.241.83)

Q4 (0.026) 1.94 (0.914.14) 1.80 (0.823.93) 0.99 (0.372.62) Q4 (0.009) 0.97 (0.422.26) 0.94 (0.392.26) 0.59 (0.211.68)

P for trend 0.02 0.03 0.69 0.95 1.00 0.37

Estradiol, pg/mL

Q1 (8) 1.09 (0.502.35) 1.10 (0.492.44) 1.95 (0.725.28) Q1 (14) 0.83 (0.332.10) 0.96 (0.362.56) 0.95 (0.322.80)

Q2 (810) 0.56 (0.261.22) 0.63 (0.281.40) 0.95 (0.362.54) Q2 (1425) 1.56 (0.663.71) 1.63 (0.673.94) 2.01 (0.785.19)

Q3 (1014) 0.81 (0.391.67) 0.85 (0.401.79) 0.73 (0.311.76) Q3 (2545) 1.54 (0.663.60) 1.66 (0.693.98) 2.13 (0.815.61)

Q4 (14) 1.00 referent 1.00 1.00 Q4 (45) 1.00 referent 1.00 1.00

P for trend 0.89 0.96 0.15 0.81 0.94 0.89

FEI

Q1 (0.0005) 0.68 (0.311.46) 0.71 (0.321.59) 1.99 (0.705.67) Q1 (0.0006) 0.99 (0.402.42) 1.07 (0.422.71) 1.47 (0.524.17)

Q2 (0.00050.0009) 0.66 (0.311.40) 0.74 (0.341.62) 1.14 (0.452.88) Q2 (0.00060.0009) 0.95 (0.392.31) 0.95 (0.382.36) 1.33 (0.503.54)

Q3 (0.00090.0015) 0.87 (0.421.82) 0.90 (0.421.94) 1.37 (0.563.30) Q3 (0.00090.002) 1.94 (0.814.63) 1.96 (0.814.76) 2.71 (1.037.15)


P for trend 0.23 0.34 0.29 0.57 0.70 0.78

FAI/FEI


Q2 (11.1215.71) 1.09 (0.502.41) 1.04 (0.462.35) 0.65 (0.251.68) Q2 (3.365.61) 0.63 (0.241.62) 0.56 (0.211.52) 0.70 (0.242.03)

Q3 (15.7121.78) 1.14 (0.522.50) 1.10 (0.492.48) 0.99 (0.392.52) Q3 (5.6111.55) 1.25 (0.532.92) 1.21 (0.512.91) 1.15 (0.452.95)

Q4 (21.78) 1.93 (0.904.13) 1.81 (0.823.99) 1.91 (0.794.61) Q4 (11.55) 1.16 (0.482.79) 1.14 (0.462.79) 1.12 (0.432.94)

P for trend 0.09 0.13 0.08 0.46 0.46 0.62

*Model 1, controlled for matching variables (age and smoking) and randomized vitamin E and aspirin.Model 2, variables in Model 1 plus family history of CHD, alcohol intake, and physical activity.Model 3, variables in Model 2 plus body mass index, history of HTN, DM, and elevated cholesterol.P for trend, trend across ordinal categories.

1692 Circulation October 7, 2003


4. Liu Y, Ding J, Bush TL, et al. Relative androgen excess and increasedcardiovascular risk after menopause: a hypothesized relation. Am J Epi-demiol. 2001;154:489494.

5. Haffner SM, Newcomb PA, Marcus PM, et al. Relation of sex hormonesand dehydroepiandrosterone sulfate (DHEA-SO4) to cardiovascular riskfactors in postmenopausal women. Am J Epidemiol. 1995;142:925934.

6. Phillips GB, Pinkernell BH, Jing TY. Relationship between serum sexhormones and coronary artery disease in postmenopausal women. Arte-rioscler Thromb Vasc Biol. 1997;17:695701.

7. Cummings SR, Browner WS, Bauer D, et al. Endogenous hormones and therisk of hip and vertebral fractures among older women: Study of OsteoporoticFractures Research Group. N Engl J Med. 1998;339:733738.

8. Hankinson SE, Willett WC, Manson JE, et al. Plasma sex steroidhormone levels and risk of breast cancer in postmenopausal women.J Natl Cancer Inst. 1998;90:12921299.

9. Barrett-Connor E, Goodman-Gruen D. Prospective study of endogenoussex hormones and fatal cardiovascular disease in postmenopausal women.BMJ. 1995;311:11931196.

10. Goodman-Gruen D, Barrett-Connor E. A prospective study of sexhormone-binding globulin and fatal cardiovascular disease in RanchoBernardo men and women. J Clin Endocrinol Metab. 1996;81:29993003.

11. Haffner SM, Moss SE, Klein BE, et al. Sex hormones and DHEA-SO4 inrelation to ischemic heart disease mortality in diabetic subjects: theWisconsin Epidemiologic Study of Diabetic Retinopathy. Diabetes Care.1996;19:10451050.

12. Rexrode KM, Lee IM, Cook NR, et al. Baseline characteristics of par-ticipants in the Womens Health Study. J Womens Health Gend BasedMed. 2000;9:1927.

13. Hankinson SE, Manson JE, Spiegelman D, et al. Reproducibility ofplasma hormone levels in postmenopausal women over a 23-year period.Cancer Epidemiol Biomarkers Prev. 1995;4:649654.

14. Wu CH, Lundy LE. Radioimmunoassay of plasma estrogens. Steroids.1971;18:91111.

15. Selby C. Sex hormone binding globulin: origin, function and clinicalsignificance. Ann Clin Biochem. 1990;27:532541.

16. Oh JY, Barrett-Connor E, Wedick NM, et al. Endogenous sex hormonesand the development of type 2 diabetes in older men and women: theRancho Bernardo study. Diabetes Care. 2002;25:5560.

17. Phillips GB, Jing TY, Laragh JH. Serum sex hormone levels in post-menopausal women with hypertension. J Hum Hypertens. 1997;11:523526.

18. Birdsall MA, Farquhar CM, White HD. Association between polycysticovaries and extent of coronary artery disease in women having cardiaccatheterization. Ann Intern Med. 1997;126:3235.

19. Golden SH, Maguire A, Ding J, et al. Endogenous postmenopausalhormones and carotid atherosclerosis: a case-control study of the athero-sclerosis risk in communities cohort. Am J Epidemiol. 2002;155:437445.

20. Bernini GP, Sgro M, Moretti A, et al. Endogenous androgens and carotidintimal-medial thickness in women. J Clin Endocrinol Metab. 1999;84:20082012.

21. Haffner SM, Valdez RA, Morales PA, et al. Decreased sex hormone-binding globulin predicts noninsulin-dependent diabetes mellitus inwomen but not in men. J Clin Endocrinol Metab. 1993;77:5660.

22. Reinecke H, Bogdanski J, Woltering A, et al. Relation of serum levels ofsex hormone binding globulin to coronary heart disease in postmeno-pausal women. Am J Cardiol. 2002;90:364.

23. Hauner H, Bognar E, Blum A. Body fat distribution and its associationwith metabolic and hormonal risk factors in women with angiographicallyassessed coronary artery disease: evidence for the presence of a metabolicsyndrome. Atherosclerosis. 1994;105:209216.

24. Lapidus L, Lindstedt G, Lundberg PA, et al. Concentrations of sex-hormone binding globulin and corticosteroid binding globulin in serum inrelation to cardiovascular risk factors and to 12-year incidence of cardio-vascular disease and overall mortality in postmenopausal women. ClinChem. 1986;32:146152.



Cook and Julie E. BuringKathryn M. Rexrode, JoAnn E. Manson, I-Min Lee, Paul M Ridker, Patrick M. Sluss, Nancy R.

Sex Hormone Levels and Risk of Cardiovascular Events in Postmenopausal Women

Print ISSN: 0009-7322. Online ISSN: 1524-4539 Copyright 2003 American Heart Association, Inc. All rights reserved.

is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231Circulation doi: 10.1161/01.CIR.0000091114.36254.F3

2003;108:1688-1693; originally published online September 15, 2003;Circulation.

http://circ.ahajournals.org/content/108/14/1688World Wide Web at:

The online version of this article, along with updated information and services, is located on the

http://circ.ahajournals.org//subscriptions/

is online at: Circulation Information about subscribing to Subscriptions:

http://www.lww.com/reprints Information about reprints can be found online at: Reprints:

document. Permissions and Rights Question and Answer this process is available in the

click Request Permissions in the middle column of the Web page under Services. Further information aboutOffice. Once the online version of the published article for which permission is being requested is located,

can be obtained via RightsLink, a service of the Copyright Clearance Center, not the EditorialCirculationin Requests for permissions to reproduce figures, tables, or portions of articles originally publishedPermissions:

http://circ.ahajournals.org/content/108/14/1688http://www.ahajournals.org/site/rights/http://www.lww.com/reprintshttp://circ.ahajournals.org//subscriptions/http://circ.ahajournals.org/

Documents

Circulation 2003 Rexrode 1688 93