The Duration of Estrogen Replacement but not Estrogen Levels Correlate with Sex Hormone Binding Globulin and Fasting Insulin.
Edward M. Lichten, M.D.1 , Denise Cunningham, R.N.1, David Pieper, Ph.D.2, Stephanie R. Lichten, B.A.3, Jason B. Lichten, M.D.4, S. LaDella, M.D.1, Jeffrey Lin, M.D.1, James Sowers, M.D.6 and the ALARM group.
AUTHORS AFFILIATIONS
1 Providence Hospital, Departments of Obstetrics & Gynecology, Southfield, Michigan
2 Providence Hospital, Department of Family Practice, Southfield, Michigan
3 St. John's Hospital, Chairman, Department of Research, Grosse Pointe, Michigan
4 Rush Medical College, Chicago, Illinois
5 Beth Israel Hospital, New York City, New York
6 S.U.N.Y.,Director, Division of Endocrinology & Metabolism, Brooklyn, New York
CORRESPONDENCE to PRINCIPLE AUTHOR
Edward M. Lichten, M.D.
Midwest Medical Group, P.C.
29355 Northwestern Highway Suite 120
Southfield, Michigan 48034
Phone (248)358-3433Fax (248)358-2513 Email:
ABSTRACT:
Background: Numerous observational studies support the cardioprotective effects fo Estrogen Replacement Therapy (ERT) on coronary risk factors. Likewise, hypo-androgenenia, elevated leels of Sex Hormone Binding Globulin (SHBG) and insulin sensitivity are associated with reduced risk of coronary disease. Although these factors seem unrelated, we hypothesize that there exists a stepwise direct correlation: 1) estrogen induces increases in SHBG, 2) SHBG decrease androgenicity, 3) increases in the duration of estrogen exposure increase SHBG and 4) duration of estrogen exposure and increases ni SHBG correlate with decreases in insulin levels. And since lower insulin levels correlate with decreases in cardiac risk, for women, the initial factor is the duration of estrogen exposure in the menopausal years.
Objective: The purpose of this cross sectional study was to determine 1) the relative impact of duration of estrogen replacement therapy (D-ERT) and venous serum Estrogens(E) levels on SHBG and insulin levels, and 2) the influence of the increases in D-ERT and SHBG on fasting insulin (i.e. insulin sensitivity) and other coronary risk factors.
Method: One hundred and sixty-three post-menopausal women who underwent routine clinic visits were evaluated by hemodynamic and anthropomorphic measures, and fasting blood sampling for plasma lipids, apoproteins, insulin, glucose, cortisol, testosterone, dehydroepiandrosterone sulfate (DHEAS), free testosterone, free estradiol, estradiol, estrone, and SHBG.
Results: One hundred and seventeen women had no history of ERT and 49 were on ERT for a duration of one to 120 months. Forty percent were African-American of whom only 10.8% were on ERT, vs whites (P > .001). There were no anthropologic differences between ERT and non-ERT cohorts in age (mean 60 + 10), presence of coronary disease, tobacco use, waist to hip ratio (WHR), cholesterol, or apo B-100. ERT was associated with a lower BMI, higher HDL-C and triglycerides (p< 0.05), and a marked increase (p<.001) in E and SHBG levels. Further, ERT was associated with a markedly lower insulin level (p<.002). There were no differences in free estradiol, testosterone, or DHEAS. The duration of estrogen replacement therapy (D-ERT) was directly related to SHBG ( r= .43 , p <.001) and inversely to fasting insulin (r = -0.51, p <.044) in the 49 women on ERT. Exclusion of the six (6) individuals on concombinant progestin therapy did not change the significance of the correlation (r = .65, p < .002). In the entire group, SHBG was inversely related to BMI, WHR, insulin, and "unbound" testosterone while it was directly correlated with HDL-C (range of p < .001 to p < .025). Estradiol levels were inversely associated in regard to age in non-estrogen users (r=-.242, p <.001) and also inversely associated with WHR in estrogen users (r= -.268, p < .006).
Conclusion: The duration of ERT (D-ERT), rather than the random estrogen measurement, was associated with increased SHBG and decreased fasting insulin. This shift in relative hormonal concentrations, perhaps mediated through increases in SHBG could possibly explain both lower fasting insulin levels and decreased CVD factors in women on long-term E replacement. SHBG represented an age-independent and lipid-independent factor that was both insulin and E dependent. The interaction between duration of Estrogen therapy and SHBG might have been the important link between sex hormones, insulin and CVD risk.
BACKGROUND
CVD is multi-factorial, and includes cholesterol, insulin and sex hormones.1 The replacement of estrogen (ERT) following menopause reduces CVD mortality by 50%2-3. Oral ERT improves risk factors for CVD including plasma lipids and fibrinogen.4 Yet, to date, researchers have been unable to relate serum estradiol levels with the presence of CVD.5-10 The duration of ERT has not been adequately researched.
The androgen sex hormone, testosterone, by itself or relative to estrogen levels, may be a relative risk factor for CVD in women.11 Exogenous testosterone may negate the benefits of ERT12, and metabolic states with high testosterone or testosterone precursors are associated with increased cardiovascular risk13, However, Haffner et al14-16 does not confirm a significant CVD risk attributable to testosterone/estrogen ratios. An indirect measurement of androgenicity, sex hormone-binding globulin (SHBG) has been reported to be directly related to insulin resistance, an atherogenic lipid profile, impaired glucose tolerance, and CVD.17-19 It is not clear, though how the high SHBG effects other CVD risk factors.
The purpose of this study has been to determine if there is an association between the duration of estrogen replacement (D-ERT) and known CVD risk factors including SHBG and fasting insulin. We have hypothesized that the duration of estrogen replacement (D-ERT) will be directly related to sex hormone-binding globulin (SHBG) and inversely related to fasting insulin levels, both recognized CVD risk factors.7,11
METHODS
Patient and Biochemical Determinations
One hundred and sixty-three postmenopausal women were identified from a multicenter CVD database in southeastern Michigan. Included for each was a comprehensive CVD risk database and anthropologic measurements of blood pressure, body mass index and waist-to-hip ratios. After documenting appropropriately informed consent, 14cc of venous blood was drawn from the antecubetal vein, spun and frozen. The serum lipid measurements were calculated at a C.D.C. approved laboratory from these specimens. A separate 2cc aliquot was used for duplicate batch hormone analysis. RAI assays were performed for estradiol, total testosterone, free testosterone, DHEAS, androstenedione, sex-hormone binding globulin (SHBG), cortisol and insulin. Estradiol, the most potent estrogen, was measured because it correlated with estrone in postmenopausal women20 and futher because its has the strongest affinity for SHBG20. The Estradiol was measured by RAI as described previously21.
The laboratory assay, technique, supplier and coefficients of variances is seen in Table I. The calculated inter- and intra-assay coefficient of variation were supplied from the supplier. The formula for calculation is C.V.= S.D./mean x 100%.
Serum cholesterol and HDL cholesterol were determined by enzymatic procedure at the C.D.C. approved laboratory in Michigan. Insulin was measured by double extraction technique.
Table I.
Laboratory AssaysCoefficients
LAB TESTS / SOURCE / MDD / INTRA-ASSAY / INTER-ASSAYCoated-tube IRA
Testosterone Total / Diagnostic(DPC) Product Corp / 2.10 ug/ml / C.V.=6.4% / C.V.=9.5%
Testosterone Free / DPC / 0.18 pg/ml / C.V.=7.3% / C.V.=9.5%
DHEA-SO4 / DPC / 2.10 ug/ml / C.V.=6.4% / C.V.=9.5%
Cortisol / DPC / 0.20 ug/dl / C.V.=4.9% / Single assay
Double Antibody Technique
Insulin / Diagnostic Systems Lab(DSL) / 1.3 mIU/ml / C.V.=4.6% / Single assay
Estradiol / DSL / 1.40 pg/ml / C.V.=7.3% / C.V.=1.3%
Androstenedione / DSL / 0.02 ng/ml / C.V.=4.5% / Single assay
“Active”
Sex Hormone Binding Globulin / Coated-tube immuno-radiometric assay (DSL) / 3 nmol/L / C.V.=7.8% / Single assay
Procedures:
Informed consent was obtained at the beginning of the examination which included measurements of height and weight. Anthropologic measurements were performed in triplicate for calculations of body mass index and waist-to-hip ratios. Blood pressure readings were taken in triplicate and averaged using a sphygmomanometer to the nearest digit on the right arm of the seated participant after at least a 5-minute rest period. Diabetes was defined as having a previous history of being treated with insulin or a hypoglycemic medication or having fasting serum glucose above 126 mg/ml. Heart disease was based on the physician's record of angina or myocardial infarction associated with changes in EKG or hospitalization/ heart catherization studies. Hypertension was based on the physician's record and the continued use of hypertensive medications. Smokers were removed from the co-variant analyses.
Although the patient's last meal was reported as the day before the evaluation, the time of day at which the blood was drawn varied from 8:00 AM to 4:30 PM. Thus, it is unlikely that the time of assay led to any systematic bias in the association between sex hormones and cardiovascular risk factors.
Statistical Analysis
All statistical analyses were performed with SPSS version 6.1. In all analyzes, a two-tailed value of P<= .05 was considered significant. In the multiple regression model, the listed variables were analyzed comparing the groups of postmenopausal women with and without estrogen replacement. In the logarithmic regression used to determine the relationship of listed variables and risk factors for cardiovascular disease, all cardiovascular risk factors were grouped together and estradiol, sex hormone binding globulin, insulin and HDL cholesterol were considered independent variables.
Although 90% of the estrogen22 and 98% of the testosterone23 in women are bound by SHBG, they were considered 100% bound for statistical analysis. Various ratios to sex-hormone binding globulin were also included in the analysis. To determine whether significant correlation existed between any two independent variables in the study, partial correlation coefficients were calculated by linear regression analysis after controlling for age and BMI.
RESULTS:
Means and standard deviations for the variables in the study appear in Table II.
Frequency of race and disease states appears in Table III. Pearson Correlation coefficients comparing variables between the group with estrogen replacement and those without estrogen appear in Table IV. Cross correlation appears in Table V. Independent and dependent variables in the equation appear in Table VI. Graphic displays of these cross group correlation appear in Scattergrams.
Figure 1 graphs the mean SHBG versus insulin for current estrogen users. Figure 2 graphs the mean SHBG versus duration of time on ERT (D-ERT). Figure 3 graphs the mean fasting insulin versus duration of time on ERT (D-ERT).
RESULTS:
The mean and standard deviations of non-estrogen users are compared to the entire group studied (Table II). There is no significant difference between the groups except in reference to the estradiol levels (E2) resulting from the supplementary estrogenreplacement. The age, BMI, waist/hip ratio, systolic and diastolic blood pressure, total testosterone, free testosterone, cortisol, DHEA-sulfate, androstenedione, and various cholesterol measurements are comparable.
The frequency table (Table III) compares the demographics of the non-estrogen and estrogen users based on race. There is a predominance of estrogen users within the Caucasian population, with 40 women reporting estrogen use versus only 7 African-American women use estrogen (p < .001). The non-ERT user population is relatively equal at 63 Caucasian versus 58 African-American women. There is a predominance of diabetes, hypertension and the all-disease-group in the non-ERT group as compared to the ERT group (p < .005).
Table IV lists the Pearson correlation coefficients of the 21 variables for the estrogen users (49) and the non-estrogen users (117) in respect to estradiol, insulin and SHBG. For both groups there is a significant inverse correlation with SHBG for insulin, BMI and Waist-hip ratio (p <. .02). There is a significant direct correlation with insulinfor BMI and Waist-hip ratio (p < .004). HDL-cholesterol is significantly and directly correlation with SHBG (p < .001) and inversely correlation with insulin (p < .034).
In non-estrogen users there exists a strong relation between SHBG and free-testosterone, triglycerides and systolic blood pressure. Triglycerides show a direct correlation and age an inverse correlation for non-estrogen users, but not in ERT users.
In the cross correlation performed between the non-estrogen and estrogen replacement groups shows that estradiol and SHBG are significantly higher and that Waist-hip ration and insulin levels are lower in the ERT group (p < .002) Table 5. In the ERT group, the BMI is lower (p < .01) while the cortisol, triglycerides and HDL-cholesterol are raised (p <.03) Table 6. Insulin is the independent variable that correlates with the presence of all disease states.
DISCUSSION
Many observational studies have shown a correlation between estrogen use and a decrease in cardiovascular disease1-3. The Framingham study1 and Nurse’s Health study2-3 showed that women who used ERT experienced dramatically less heart disease. However, measurements of random estradiol, total estrogen, and free estradiol have not show significant associations with cardiovascular risk parameters.5-9 One of the directives of the Women's Health Initiative begun in 1992 was to determine the role that ERT might have in reducing cardiovascular risk in menopausal women.
The current investigation is the first to incorporate the parameter of duration of time on estrogen (D-ERT) within a post-menopausal data base. Previous reports have focused on the inter-correlation of three cardiovascular risk factors for post-menopausal women: SHBG, insulin, and cardiovascular disease.7,11,14-16,19
SHBG serves as the important determinant of the ratio of unbound estradiol and unbound testosterone. In biological female systems, increases in SHBG have a feminizing effect while decreases in SHBG are masculinizing.20 In addition to being directly effected by the level of various sex hormones, SHBG is also strongly influenced by states of hyperinsulinemia and obesity21. Hyperinsulinemia with and without obesity prove to be masculinizing for the polycystic female.2 SHBG and insulin are both recognized independent risk factor of cardiovascular disease,22, 23 independent of HDL cholesterol, triglycerides, Apolipoprotein-B, and HDL-C/ Cholesterol ratio. In this complex system, SHBG may serve as the primary conduit of the action of sex hormones on insulin and insulin, in turn, on sex hormones.
In conclusion, the Duration of Estrogen Replacement (D-ERT), rather than any random measurement of any estrogenic component was associated with increased SHBG and decreased fasting insulin. Although the domino effect of increased D-ERT contributes to increases in SHBG and decreases in androgenicity, it is their effect on insulin that is primary to reduction in risk. For the logarithmic regression analysis confirms that insulin remains the only independent contributor to disease risk.
Since D-ERT induces changes in SHBG that acts on the cell wall to facilitate transport of hormones into the cell, further research may discover a complementary effect on the cell wall action of glucose transport.
We conclude that in postmenopausal women, it may well be the lack of long-term estrogen replacement and decreased SHBG that define a state of increased fasting insulin and cardiovascular risk. When future prospective studies are completed that compare estrogen preparations, SHBG, insulin and cardiovascular risk, we may have proof that decreased cardiovascular risk is observed in those individuals who remain the longest on estrogen replacement.
TABLE II: POSTMENOPAUSAL WOMEN:
Mean and Standard Deviations
ALL WOMENNON-ESTROGEN USERS
variableMeanStandard Dev Cases MeanStandard Dev Cases
Insulin 11.2334 7.7102160 | 12.4113 8.0348116
SHBG143.077687.3985161 |115.025660.3077117
E2 43.844464.6287162 | 19.515441.2115117
E2 x SHBG 8537.3315723.9161 |2390.1716193.02117
Testosterone 0.2831 0.3604162 | 0.2625 0.2103117
Testosterone/SHBG 0.0027 0.0032161 | 0.0030 0.0028117
T-free 0.9183 0.8377161 | 0.8991 0.6139117
Cortisol 13.5725 5.8470160 | 12.8276 5.0179116
DHEA-S101.219059.8350162 | 101.862 58.369117
Androstenedione 0.8275 0.4208160 | 0.8492 0.4203117
Triglycerides127.472470.5147163 |120.092467.6587119
Total Cholesterol229.349639.8326163 |226.974742.8289119
HDL Cholesterol 55.331214.9166163 | 53.756314.1613119
LDL/Cholesterol153.944837.1552163 |154.462240.4271119
HDL-C/ Chol 0.2480 0.0076163 | 0.2449 0.0078119
Apoprotein(B)115.466229.1764163 |115.042030.7341119
BP- Systolic132.5195 18.6205145 |133.077718.4576103
BP- Diastolic 80.8069 9.7616145 | 81.440110.1131103
Age 59.96 10.96163 | 60.3811.36125
BMI 30.3031 6.8116160 | 31.105811.36115
Waist/Hip Ratio 0.8220 0.0735161 | 0.8310 0.0074125
======
Table III.Frequency Table
Population ERT non-ERT p value
Women (172) / 49 / 117 / **Race
~White / 40 (38.8%) / 63 (61.2%) / 0.001**
~Black / 7 (10.8%) / 58 (89.2%)
Diabetes (172) / 1 ( 2.1%) / 19 (15.6%) / 0.005**
Smoker (171) / 7 (15.0%) / 13 (10.5%) / n.s.
Angina (169) / 3 ( 6.4%) / 17 (13.9%) / n.s.
Hypertension (169) / 15 (31.9%) / 61 (50.0%) / 0.025*
MI (169) / 1 ( 2.1%) / 4 ( 3.3%) / n.s.
All diseases (169) / 16 (34.0%) / 72 (59.0%) / 0.003**
**Power .80 @ .05 Alpha
Table V.
Pearson Correlation coefficients: Cross Groups
ERTNo ERTp value
Number of women / 49 / 117Age / 57.8 9.7 / 60.4 11.4 / n.s.
Race
~White / 40 (38.8%) / 63 (61.2%) / .001
~Black / 7 (10.8%) / 58 (89.2%)
SHBG (49) / 217 + 104 / 115 + 60 / .001
Estradiol / 107 + 72 / 19.5 + 41.2 / .001
Insulin / 8.12 + 5.8 / 12.4 + 8.0 / .002
Cortisol / 15.5 + 7.3 / 12.8 + 5.0 / .01
BMI / 28.3 + 5.7 / 31.1 + 7.0 / .02
Triglycerides / 147 75 / 120 68 / .03
HDL Cholesterol / 60.0 16 / 53.7 14 / .03
DHEAS / 99.5 64 / 102 58 / n.s.
Testosterone / 0.337 0.59 / 0.262 0.21 / n.s.
Free Estradiol / 0.722 0.61 / 0.560 0.50 / n.s.
Total Cholesterol / 235 30 / 227 43 / n.s.
Chol/HDL ratio / 4.22 1.2 / 4.46 1.3 / n.s.
HDL/Chol ratio / 0.256 0.073 / 0.245 0.078 / n.s.
Apoprotein B / 117 25 / 115 31 / n.s.
Waist to Hip / 0.79 + 0.067 / 0.83 + 0.074 / .008
VI. Dependent variable .. DISEASE (MI, HTN, DM, ANG)
-2 Log likelihood214.55939
- constant is included in the model
Variables entered on Step Number
1..Insulin
HDLCHOL
SINAIHDL
E2
Testosterone
SHBG
Estimation terminated at iternation number 4 because Log likelihood decreased by less than .01 percent
-2 Log likelihood189.698
Goodness of fit172.253
Cox and Snell –R^2 .148
Natgelkerke -R^2 .198
Chi QuaredrSignificance
Mdoel24.8626.0004
Block24.8626.0004
Step24.8626.0004
Classification Table for DISEASE
The cut value is .50
Predicted
NoYesPercent correct
Observed No502467.57%
Observed Yes285365.43%
Overall66.45%
Variables in the Equation: Logrithmic Regression
VariableB S.E.WalddfSig R Exp(B)
Insulin / 0.0898 / 0.0290 / 9.5883 / 1 / 0.0020 / 0.1892 / 1.0940HDL/C / -1.2285 / 3.5175 / 0.1220 / 1 / 0.7269 / 0.0000 / 0.2927
HDL / 0.0154 / 0.0184 / 0.7033 / 1 / 0.4017 / 0.0000 / 1.0155
Estradiol / -0.0049 / 0.0029 / 2.8997 / 1 / 0.0886 / -.0658 / 0.9981
Testosterone / -0.7494 / 0.7129 / 1.1049 / 1 / 0.2932 / 0.0000 / 0.4727
SHBG / -0.0028 / 0.0025 / 1.2267 / 1 / 0.2680 / 0.0000 / 0.9972
Constants / -1.0630 / 0.8682 / 1.1049 / 1 / 0.2208
KEYWORDS
sex hormone binding globulin, testosterone, estradiol, insulin, Syndrome-X, cholesterol, obesity, dehydroepiandrosterone, risk factors, coronary artery disease, women, correlation
Abbreviations:
ApoB=Apoprotein-(B)
BMI=body mass index
CAD=coronary artery disease
ASHD=atherosclerotic heart disease
F=cortisol
DHEAS=Dehydroepiandrosterone sulfate
DPC=Diagnostic Products Corp
DSL=Diagnostic Systems Lab
FT=free testosterone
TRT=Testosterone Ratio Test
HDL/Chol =HDL/ Cholesterol ratio
HDL=high density lipoprotein cholesterol
LDL=low density lipoprotein cholesterol
MDD=minimal detectable dose
Tri=triglycerides
RAI=radioimunoassay
SHBG =sex-hormone-binding globulin
TC=total cholesterol
TRT=Testosterone Ratio Test
TT=total testosterone
WHR=waist-to-hip ratio
Acknowledgments:
Quest Laboratories, Capistrano, California for funding the free estradiol and total estrogen assays.
Lori Mosca, M.D, Ph.D. who supplied the information from the A.L.A.R.M. database and initial support for this study.
Grant supported was supplied by Providence Hospital, Research Fund. We are grateful to Kathleen Lobocki, M.T. who performed all of the non-lipid analyzes in the Research Department Laboratories of Providence Hospital, Southfield, Michigan.
References
1Kannel WB, Hjortland MC, McNamara PM, Gordon T. Menopause and risk of cardiovascular disease. The Framingham Study. Ann Intern Med 1976;85(4):447-52
2Stampfer MJ, Colditz GA, Willett WC, Manson JE, et al. Postmenopausal estrogen therapy and cardio-vascular disease: Ten year follow-up from the nurses health study. N Engl J Med1991;325(11):756-62