Baseline Serum Estradiol and Fracture Reduction during Treatment with Hormone Therapy: The Women's Health Initiative Randomized Trial Jane A. Cauley, DrPH Andrea Z. LaCroix, PhD John A. Robbins, MD Joseph Larson, MS Robert Wallace, MD Jean Wactawski-Wende, PhD Zhao Chen, PhD Douglas C. Bauer, MD Steven R. Cummings, MD Rebecca Jackson, MD Osteoporos Int 2010 Jan;21(1):

Hypotheses Primary Hormone therapy decreases risk of fracture to a greater degree among women with the lowest circulating estrogen levels. Secondary Women with lowest baseline estradiol will have a higher risk of fracture.

Design WHI E+P and E-alone trials. Nested case-control study. All confirmed cases of hip fracture (n=248) & random sample of other fx until 750 pairs. Controls matched on age, ethnicity, HT trial RV date, hysterectomy status and past hx Fx

Design Matched first on hip fracture then on other fractures – Age (± 1 yr) – Ethnicity – Hormone therapy randomization date (± 1 yr) – Prior history of fracture – Hysterectomy status 231 Hip fracture pairs 519 Random sample of other fracture pairs 750 Total Pairs

Sex Steroid Hormones Reproductive Endocrine Research Laboratory (USC) Estradiol: extraction/chromatography RIA – Sensitivity, 3 pg/ml (11.0 pmol/L) – Intra-assay CV: 7.9%; Inter-assay CV: 8-12% Free and bioavailable: calculated SHBG: solid phase, 2 site chemiluminescent immunoassay – Sensitivity, 0.2 nmol/L – Intra-assay CV: 4-7%; Inter-assay CV: %

Statistical Analyses T-tests chi-square: compare characteristics of women by randomized group and fracture status. Interaction between RX group x biomarkers. Unconditional logistic regression to test whether biomarker predicted fracture risk.

Baseline Biomarkers by Case-control Status & by Randomized Group: All Fracture CasesControls E+P, E (n=304) PBO (n=446)P Value* E+P, E (n=304) PBO (n=333)P Value* P Value** Mean Total Estradiol (pg/mL) Free Estradiol (pg/mL) Bio E2 (pg/mL) SHBG (μg/dL) *P value from a linear model modeling the biomarker of interest as a function of HT status. ** P value from a linear model modeling the biomarker of interest as a function of cases-control status.

Baseline Biomarkers by Case-Control Status & by Randomized Group: Hip Fracture CasesControls E+P, E (n=95) PBO (n=136)P Value* E+P, E (n=134) PBO (n=97)P Value* P Value** Mean Total Estradiol (pg/mL) Free Estradiol (pg/mL) Bio E2 (pg/mL) SHBG (μg/dL) < *P value from a linear model modeling the biomarker of interest as a function of HT status. ** P value from a linear model modeling the biomarker of interest as a function of cases-control status.

All fracturesHip fractures E+P, EPBOOR 95% CI for fracture Inter- action P- Value E+P, EPBOOR 95% CI for fracture Inter- action P- Value Total Estradiol (pg/mL) <6 < (0.33, 0.74) (0.17, 0.79) > (0.33, 0.75) (0.30, 1.30) > (0.32, 0.83) (0.18, 1.06) > (0.36, 0.85) (0.23, 1.33) Adjusted for age, ethnicity, randomization date, fracture history and hysterectomy status. OR (95% CI) of Fracture among Women Randomized to HT compared to PBO according to Levels of Total Estradiol (pg/mL) at Baseline

All fracturesHip fractures E+P, EPBOOR 95% CI for fracture Inter- action P- Value E+P, E PBOOR 95% CI for fracture Inter- action P- Value <4< (0.34, 0.74) (0.20, 0.83) > (0.28, 0.74) (0.20, 1.06) > (0.36, 0.84) (0.34, 1.57) > (0.35, 0.85) (0.20, 1.28) Adjusted for age, ethnicity, randomization date, fracture history and hysterectomy status. OR (95% CI) of Fracture among Women Randomized to HT compared to PBO according to Levels of Bioavailable Estradiol ( pg/mL) at Baseline

All fracturesHip fractures E+P, EPBOOR 95% CI for fracture Inter- action P- Value E+P, EPBOOR 95% CI for fracture Inter- action P- Value SHBG (μg/dL) < (0.35, 0.84) (0.12, 1.03) > (0.52, 1.17) (0.45, 2.15) > (0.31, 0.72) (0.29, 1.23) > (0.27, 0.61) (0.15, 0.64) Adjusted for age, ethnicity, randomization date, fracture history and hysterectomy status. OR (95% CI) of Fracture among Women Randomized to HT compared to PBO according to Levels of SHBG (μg/dL) at Baseline

OR (95% CI) of Fracture among Women Randomized to Hormone Therapy compared to Placebo According to Levels of Sex Steroid Hormones at Baseline E+P vs. E alone Trial E+PE alone trial E casesPBO cases OR 95% CI for fracture Inter- action P- Value E cases PBO cases OR 95% CI for fracture Inter- action P- Value Bioavailable E < (0.37, 1.05) (0.22, 0.70) > (0.17, 0.65) (0.30, 1.29) > (0.27, 0.87) (0.34, 1.23) > (0.24, 0.77) (0.39, 1.47) Adjusted for age, ethnicity, randomization date, fracture history and hysterectomy status.

Hypotheses Primary Hormone therapy decreases risk of fracture to a greater degree among women with the lowest circulating estrogen levels. Secondary Women with lowest baseline estradiol will have a higher risk of fracture.

Odds Ratio (95% CI) of Fracture according to Levels of Bioavailable E2 at Baseline: All Fractures Full MV model OR 95% CI for fracture P trend Bioavailable Estradiol (pg/mL) <4<41.00 > (0.56, 1.04) > (0.70, 1.25) > (0.66, 1.24) Adjusted for age, ethnicity, randomization date, fracture history and hysterectomy status, BMI, treated diabetes and self-reported health.

Odds Ratio (95% CI) of Fracture according to Levels of BIOE2 at Baseline: Hip Fractures Main Effects Model 1 Base Model* OR 95% CI for fracture P trend Bio E2 (pg/mL)0.056 <4 <41.00 > (0.33, 0.94) (0.40, 1.08) > (0.28, 0.89) * Base model adjusted for age, ethnicity, randomization date, fracture history and hysterectomy status. Additional adjustment for BMI, treated diabetes and self- reported health. Model 2 Adjusted for Base Model** OR 95% CI for fracture P trend (0.37, 1.08) 0.80 (0.47, 1.35) 0.75 (0.40, 1.39) Cauley JA et al. Osteoporos Int 2010;21:

Odds Ratio (95% CI) of Fracture according to Levels of SHBG at Baseline: All Fractures Model 3 Full MV model OR 95% CI for fracture P trend SHBG (μg/dL) < > (0.95, 1.74) > (0.82, 1.53) > (1.11, 2.12) Adjusted for age, ethnicity, randomization date, fracture history and hysterectomy status, BMI, treated diabetes and self-reported health.

Odds Ratio (95% CI) of Fracture according to Levels of SHBG at Baseline : ALL Fractures and Hip Fractures ALL Fractures OR 95% CI for fracture P trend SHBG (μg/dL) < > (0.95,1.74) > (0.82, 1.58) > (1.11, 2.12) * Base model adjusted for age, ethnicity, randomization date, fracture history and hysterectomy status. Additional adjustment for BMI, treated diabetes and self- reported health. Hip Fx OR 95% CI for fracture P trend (1.08, 4.14) 2.69 (1.37, 5.39) 4.01 (1.98, 8.15)

Summary – The effect of HT on fracture reduction is independent of Baseline E2 and SHBG. Similar results for E&P and E alone – There was no association between E2 and all fx. Women with higher E2 at baseline had a lower risk of Hip fx but this was attenuated in models with BMI. – Women with higher levels of SHBG have a higher risk of All Fx and specifically hip fractures. Independent of BMI and other confounding variables and Bio E2

Sex Steroid Hormones and Fracture in Multi-Ethnic Women – Nested case Control design: WHI OS – All fractures in Blacks, Hispanics, Asians and Native Americans; Random sample of 400 White women with FX – Cases and Controls matched on age, race/ethnicity and date of randomization – USC Endocrine lab – Poster this evening

MV Adjusted Odds Ratio (95% CI) of Clinical Fractures Across Categories of 25(OH)D: WHI Cauley JA, et al. J Bone Miner Res 2011 Adjusted for age, blood draw date, clinic, height, weight, physical activity, calcium intake, history of fractures N cases N controls

Tertiles P trend 123 Whites 1,2,3 Ref0.71 (0.48, 1.05)0.52 (0.33, 0.80)0.003 Blacks 1,2,3 Ref0.65 (0.40, 1.04)0.52 (0.32, 0.86)0.012 Hispanics 1,2,3 Ref1.33 (0.68, 2.61)1.08 (0.52, 2.23)0.86 Asians 1,2,3 Ref0.74 (0.38, 1.47)1.01 (0.44, 2.30)0.88 Native Americans 1,2,3 Ref0.36 (0.05, 2.87)0.22 (0.03, 1.54) Matched on age, race, date of blood draw. 2 MV model: adjusted for weight, height, physical activity, calcium intake, history of fracture. 3 Tertile cutoffs = , >8.21. Odds Ratio (95% CI) of Fracture across Tertiles of Bioavailable Estradiol (pg/ml)

Tertiles P trend 123 Whites 1,2,3 Ref 1.02 (0.76, 1.46) 0.79 (0.54, 1.16) 0.25 Blacks 1,2,3 Ref 0.90 (0.57, 1.40) 0.54 (0.34, 0.85) Hispanics 1,2,3 Ref 1.04 (0.58, 1.87) 1.02 (0.54, 1.91) 0.95 Asians 1,2,3 Ref 0.89 (0.47, 1.67) 0.64 (0.28, 1.49) 0.32 Native Americans 1,2,3 Ref 0.34 (0.06, 1.90) 0.10 (0.01, 0.81) Matched on age, race, date of blood draw. 2 MV model: adjusted for weight, height, physical activity, calcium intake, history of fracture. 3 Tertile cutoffs = , >13.4 Odds Ratio (95% CI) of Fracture across Tertiles of Bioavailable Testosterone (pg/ml)

Tertiles P trend 123 Whites 1,2,3 Ref 0.81 (0.56, 1.16) 0.65 (0.44, 0.96) Blacks 1,2,3 Ref 1.60 (1.08, 2.37) 1.51 (0.98, 2.32) Hispanics 1,2,3 Ref 1.04 (0.57, 1.90) 0.90 (0.46, 1.76) 0.81 Asians 1,2,3 Ref 2.14 (0.90, 5.08) 0.84 (0.37, 1.95) 0.74 Native Americans 1,2,3 Ref 1.36 (0.38, 4.81) 0.56 (0.12, 2.48) Matched on age, race, date of blood draw. 2 MV model: adjusted for weight, height, physical activity, calcium intake, history of fracture. 3 Tertile cutoffs = , >58.2. Odds Ratio (95% CI) of Fracture across Tertiles of Sex Hormone Binding Globulin (nmol/L)

Endogenous Estrogen Levels and CHD Outcomes in the WHI Trials DC Bauer, G Farhat, JA Cauley, A Huang, A LaCroix, J Lee, D Grady, K Yaffe, J Manson, N Parimi, E Vittinghof and SR Cummings for the Women’s Health Initiative (WHI) Research Group Coordinating Center

Research Questions Among women randomized to E-alone or E+P in the WHI: Are pre-randomization estradiol levels associated with CHD risk? Do pre-randomization estradiol levels modify the effects of either E-alone or E+P on CHD risk?

Pre-Treatment Estradiol and SHBG by CHD Status † Median (IQR) * p<0.05 compared to no CHD E+P Trial E-alone Trial CHD No CHD CHD Total E2 (ng/ml) Total E2 (ng/ml) † 10.2* ( ) 11.0 ( ) 10.8 ( ) 10.6 ( ) Bioavail E2 (ng/ml) Bioavail E2 (ng/ml) † 6.5 ( ) 7.0 ( ) 7.1 ( ) 7.2 ( ) SHBG (nmol/L) 43.8 ( ) 43.8 ( ) 42 ( ) 38.3 ( ) 38 ( )

Adjusted CHD Risk by Quartile of Total E2 (pg/ml) for E-alone Trial Q1 (ref) Q2 Q 3 Q4 Hazard Ratio 95% CI (2-7.5) ( ) ( ) ( ) P for Trend = 0.019

Adjusted CHD Risk by Quartile of Total E2 (pg/ml) for E+P Trial Q1 (ref) Q2 Q 3 Q4 Hazard Ratio 95% CI ( ) ( ) ( ) ( ) P for Trend =

Adjusted CHD Risk by Quartile of Bio-available E2 (pg/ml) for E+P Trial Q1 (ref) Q2 Q3 Q4 Hazard Ratio 95% CI ( ) ( ) ( ) ( ) P for Trend = 0.007

Adjusted CHD Risk by Quartile of Bio-available E2 (pg/ml) for E-alone Trial Q1 (ref) Q2 Q 3 Q4 Hazard Ratio 95% CI ( ) ( ) ( ) ( ) P for Trend = 0.011

Adjusted* CHD Risk in Lowest Quartile Vs. Other Three Quartiles RH (95% CI) E-alone E+ P Total E2 1.6 ( ) 2.8 ( ) Bio- available E2 2.1 ( 1.2 – 3.8) 1.8 ( ) SHBG 0.8 ( ) 1.4 ( ) *Adjusted for treatment, age, education, time since menopause, current smoking, alcohol, waist circumference, prior HT use, prior oral contraceptive use, high blood pressure, diabetes, high cholesterol, history of heart disease, aspirin use, and statin use

Research Questions Among women randomized to E alone or E + P in the WHI: Are pre-randomization estradiol levels associated with CHD risk? Do pre-randomization estradiol levels modify the effects of either E-alone or E+P on CHD risk?

Pre-randomization E2 and Effect of Treatment on CHD Risk E-alone vs. placebo – Risk of CHD independent of E2 levels – Interaction p-value=0.18 E+P vs. placebo – Risk of CHD independent of E2 levels – Interaction p-value=0.73

Summary Among postmenopausal women enrolled in the WHI trials – Higher endogenous E2 associated with reduced risk of CHD – Independent of many known CV risk factors – SHBG not associated with CHD risk No evidence that the effects of E-alone or E+P on CHD risk differ by pre- randomization estradiol levels

Bio E2 Stratified by Median Time Since Menopause for the E-alone Trial Time since menopause > 22 yrs P for trend = Time since menopause <=22 yrs P for trend= Hazard Ratio 95% CI Q1 Q2 Q3 Q4 Interaction p value = 0.22