Influence of Cardiometabolic Status on Cardiovascular Effects of Oral Menopausal Hormone Therapy

Precis:

Blood lipids could assist with selection of patients for menopausal hormone therapy.

Objective:

To assess risk of cardiovascular outcomes from menopausal hormone therapy (MHT) by cardiometabolic status.

Methods:

Secondary analysis of 2 double-blind placebo-controlled randomized controlled trials of conjugated equine estrogens (CEE, 0.625 mg/d) or CEE with medroxyprogesterone acetate (MPA, 2.5 mg/d) versus placebo in postmenopausal women aged 50–79. The primary outcome was coronary heart disease (CHD, non-fatal myocardial infarction or CHD death). Cardiometabolic status was evaluated by lipid profile, blood pressure, blood glucose, and presence of metabolic syndrome.

Results:

The CEE-alone trial enrolled 10,739 participants with a hysterectomy and the CEE+MPA trial enrolled 16,608 participants with an intact uterus. Randomization to oral MHT did not increase CHD risk in participants with a history of treated hyperlipidemia, or in untreated participants with favorable lipid profiles, but risk increased in those with unfavorable lipid profiles. For CEE+MPA the CHD risk in untreated participants with normal LDL cholesterol (<130 mg/dl) was similar to placebo (HR 0.59; 95%CI: 0.31–1.10); however, for elevated LDL ≥190mg/dl, the risks were more than doubled (HR 2.77; 95%CI: 1.42–5.40 (P-trend=0.002). CHD risks increased with increasing LDL/HDL ratio (ratio <2.5 HR 0.73; 95%CI: 0.39–1.37 compared to ratio ≥4 HR 1.76; 95%CI: 1.08–2.88 (P-trend=0.008). A similar but non-significant pattern of risk by increasing level of LDL/HDL ratio was seen for CEE-alone versus placebo. Participants with HDL cholesterol levels ≥60 mg/dl compared to <50 mg/dl appeared to be at somewhat lower risk (HRs 0.68 and 1.24, P-trend=0.02).

The patterns for effect modification on CHD risk by untreated LDL cholesterol levels within 10-year age groups (50–59, 60–69, and 70–79 years) were consistent with the overall results in both trials. Blood pressure, blood glucose, triglycerides, or metabolic syndrome did not modify CHD risk due to CEE or CEE+MPA.

Conclusion:

Higher baseline levels of LDL cholesterol when using CEE+MPA or lower levels of HDL cholesterol when using CEE showed increased risk for CHD across all age groups, and prior treatment of hyperlipidemia demonstrated no increased risk of CHD on CEE+MPA or CEE compared to placebo. Blood lipids could assist with selection of patients for CEE and CEE/MPA use.

Clinical trial registration:

ClinicalTrials.gov, https://clinicaltrials.gov, NCT0000611.

INTRODUCTION:

Removal of the “black box” warning for menopausal hormone therapy (MHT) may increase the use of MHT in younger women for treatment of vasomotor symptoms (VMS).¹ Some guidelines suggest that treatment may be extended into older ages on an individualized basis.² However, initiation of oral MHT in older women with moderate or severe VMS puts them at high risk for CHD.³ In addition to no overall benefit for CHD prevention, clinical trials showed an increase in venous thromboembolism (VTE, deep vein thrombosis or pulmonary embolism) and stroke, and an increased risk of CHD during the first years following randomization.^4–13 Risk stratification including routinely measured treatable factors that modify the risk of cardiovascular diseases (CVD) due to MHT could assist clinicians in tailoring individual therapy.²

MHT effects by baseline biochemical and physical variables were prespecified in the protocol of the Women’s Health Initiative (WHI).¹⁴ Prior WHI studies indicated that women with elevated low-density lipoprotein (LDL) cholesterol levels, high LDL/high-density lipoprotein (HDL) ratios, or with metabolic syndrome (MetS) experienced increased risk of CHD on oral MHT treatment.^15–18 However, these studies had incomplete biomarker data from the first 4 years of the trials only.

The focus in this report is on modification of the effect of oral MHT on CVD stratified by treatment of cardiometabolic risk factor, and by level of cardiometabolic risk factor in untreated participants. We leveraged the now-available, more complete biomarker data and applied more efficient statistical techniques to examine effects on a wider range of outcomes (CHD, stroke, all-cause mortality, and VTE) over the full WHI trial periods. We also examined effect modification during the initial 2 years following randomization when risk of CHD due to oral MHT is higher, and the effects by 10-year age groups.

METHODS

During enrollment from 1993–1998 at 40 US clinical centers postmenopausal women aged 50–79 years at baseline were randomized 1:1 to conjugated equine estrogens (CEE)-alone (CEE 0.625 mg/d) versus placebo if they had a hysterectomy (N=10,739) or to CEE+ medroxyprogesterone (MPA 2.5 mg/d) versus placebo if they had a uterus (N= 16,608) with planned trial durations of 8.4 years. To enhance generalizability enrollment approximated the self-reported race and ethnic proportions in the US population of women of this age group. The double-blind parallel study design, baseline characteristics of participants, methods of recruitment and enrollment criteria have been published previously.¹⁹ The trial of CEE+MPA was stopped early (at median follow up of 5.2 years, full follow-up 5.6 years) because of increased risks of CHD (the primary outcome), stroke, pulmonary embolism, and breast cancer, while the CEE-alone trial was stopped early (at median follow up of 6.8 years, full follow up 7.2 years) because of an increased risk of stroke and unlikelihood that benefit for CHD would be shown.^5–7 The trials were approved by Institutional Review Boards at 40 US clinical centers. Cardiometabolic status was evaluated by lipid profile, blood pressure, blood glucose, and presence of metabolic syndrome.

At baseline, fasting blood samples were collected, processed and stored at −80°C in a central biorepository. Assays were run at 2 central laboratories 2–15 years after collection. Because of a change in central laboratory and simultaneous change from plasma to serum for lipid measurements, we harmonized data from the 2 laboratories and recalibrated plasma lipids to serum values using 3,118 paired plasma/serum samples. We recalculated LDL values using the Sampson equation which allows for estimation of LDL cholesterol in subjects with hypertriglyceridemia and is more accurate at lower LDL levels than the older Friedewald equation.^20,21 Assessment of the presence or absence of MetS at the baseline visit required at least three of five (Adult Treatment Panel III [ATPIII] National Cholesterol Education Program 2004) criteria: waist size >88 cm, systolic blood pressure >130 mm Hg or diastolic blood pressure >85 mm Hg (or hypertension), fasting glucose >100 mg/dL (or diabetes), HDL cholesterol <50 mg/dL, or triglycerides >150 mg/dL.²² For each risk factor, we analyzed participants reporting treatment for that factor or taking medications based on an in-person inventory conducted by trained staff as a separate group, so that the unconfounded influence of the untreated risk factor can be examined in the remaining women. Detailed methods for physical assessment and biomarker measurements are described elsewhere.^5,6,15–18

The primary outcome was CHD (non-fatal myocardial infarction or CHD death). The secondary outcomes were stroke, all-cause mortality, and VTE. Methods of ascertainment and physician adjudication of clinical events have been published previously.²³

Event rate comparisons of MHT versus placebo were based on the intent-to-treat principle using failure time methods in a full-cohort analysis. For a given outcome, the time to event was the number of days from randomization to the first diagnosis of the designated event. Comparisons of outcomes included number of events with annualized rates (total number of events divided by total follow-up time in years and expressed as a percentage), differences in absolute risks per 10,000 person-years, and hazard ratios (HRs) with 95% confidence intervals (CIs). Secondary analyses examined effects in the first 2 years after randomization, and by 10-year age groups (50–59, 60–69, and 70–79 years).

Biomarkers were aggregated from several sub-studies, resulting in 54.7% of participants in the CEE-alone trial having complete biomarker data (including 89.2% of CHD cases) and 49.2% in the CEE+MPA trial (including in 94.1% of CHD cases). For participants not selected for sub-studies, biomarkers were multiply imputed (M=50) under a missing-at-random assumption, as missingness was primarily due to study design (Figure 1, details of multiple imputation methods in Appendix 1, available online at http://links.lww.com/xxx).^24,25 In time-to-event data, most statistical information resides in the cases, as the variability of the estimated log-HR is inversely proportional to the number of cases, therefore the uncertainty attributable to missing data is small. Cox-regression models included covariate adjustment for age (linear), self-reported race and ethnicity, prior MHT use, and history of CVD, and stratified the baseline hazard function by age group, dietary modification trial participation, current VMS, and history of relevant prior disease. Full-cohort analyses with imputed data included the main effect of CEE-alone vs placebo and CEE+MPA vs placebo in all women that were randomized. To evaluate effect modification by treatment status and level of risk factor in untreated participants, the core model was expanded to include a treatment × subgroup interaction term, with subgroup coded as an integer for ordered levels, and additional stratification of the baseline hazard by subgroup; the resulting 1-degree-of-freedom interaction test was reported as P-trend. Rubin’s rules were applied to combine Cox regression estimates across imputed datasets;²⁴ 2-sided P values less than .05 were regarded as statistically significant. A sensitivity analysis restricted to participants with complete biomarker data was conducted (Appendix 2, available online at http://links.lww.com/xxx). All statistical analyses were conducted using SAS software version 9.4 (SAS Institute Inc, Cary, North Carolina) and R software version 4.4 (R Foundation for Statistical Computing).

Figure 1.

Participant flow diagram. Menopausal hormone therapy (MHT) trial participants were included in several Women’s Health Initiative (WHI) biomarker substudies, resulting in approximately half having complete biomarker data, including most participants with incident coronary heart disease (CHD) (primary endpoint). Biomarkers for remaining participants were imputed. *Stratified random sample with strata defined by clinical center (n=40), hysterectomy status, age group, and 6-fixed race and ethnicity categories (self-reported). †Census of all American Indians and Alaska Natives (self-reported) providing supplemental consent. ^‡Virtual census of Non-Hispanic Black women and Hispanic women (self-reported) providing supplemental consent. ^§Stratified random sample of Non-Hispanic White women (self-reported) with strata defined by hysterectomy status, age group, incidence and history of CHD, stroke, venous thromboembolism (VTE), and diabetes. ^ǁCase-control study of all CHD, stroke, and VTE cases through 2001 and matched controls; matched on age, hysterectomy status, and disease history. CEE, conjugated equine estrogens; MPA, medroxyprogesterone acetate; W, core WHI study; A, ancillary study; NA, Native American; EA, European American.

RESULTS

The mean age of participants in the CEE-alone and CEE+MPA trials were 63.6 and 63.3 years, respectively. Proportions of participants with a history of treated hyperlipidemia, hypertension, or diabetes were similar between randomization groups within each trial (absolute differences ≤ 1.1%), but more prevalent in the CEE-alone than CEE+MPA trial (e.g., treated diabetes prevalence, 8.0%–8.1% vs 4.6%–4.7%, respectively; Table 1). Approximately 15% of CEE-alone and 12% of CEE+MPA participants had a history of treated hyperlipidemia, about half of which was with statins. Overall, cardiometabolic health profiles were also similar between randomization groups, and uniformly less favorable in the CEE-alone trial than the CEE+MPA trial (Appendix 3, available online at http://links.lww.com/xxx). Additionally, participants previously treated for CVD risk factors had less favorable profiles than untreated participants (e.g., glucose concentration, 166–174 vs 94–95 mg/dL).

Table 1.

Participant Characteristics Collected at Baseline

	CEE-alone Trial				CEE+MPA Trial
Characteristic	CEE-alone (n=5310)		Placebo (n=5429)		CEE+MPA (n=8506)		Placebo (n=8102)
	N*	%	N*	%	N*	%	N*	%
Age at screening, y, mean (SD)	63.6	(7.3)	63.6	(7.3)	63.2	(7.1)	63.3	(7.1)
Age group at screening, y, %
50–59	1639	30.9	1674	30.8	2837	33.4	2683	33.1
60–69	2386	44.9	2465	45.4	3854	45.3	3655	45.1
70–79	1285	24.2	1290	23.8	1815	21.3	1764	21.8
Race and ethnicity†, %
American Indian	41	0.8	34	0.6	25	0.3	30	0.4
Asian/Pacific Islander	86	1.6	78	1.4	194	2.3	169	2.1
Black	781	14.7	835	15.4	548	6.4	574	7.1
White	4009	75.5	4075	75.1	7141	84.0	6805	84.1
Hispanic	319	6.0	332	6.1	471	5.5	415	5.1
None of the above	74	1.4	75	1.4	127	1.5	109	1.3
Prior MHT use‡, %	2541	47.9	2659	49.0	2229	26.2	2080	25.7
Moderate or severe VMS, %	913	17.4	917	17.1	1072	12.7	974	12.1
Smoking, %
Never	2723	51.9	2705	50.4	4178	49.6	3999	50.0
Past	1986	37.8	2090	38.9	3362	39.9	3157	39.5
Current	542	10.3	571	10.6	880	10.5	838	10.5
Statin use	397	7.5	430	7.9	580	6.8	535	6.6
History of treated hyperlipidemia	763	14.4	829	15.3	1018	12.0	1027	12.7
History of treated hypertension	2202	41.5	2195	40.4	2512	29.5	2423	29.9
History of treated diabetes	432	8.1	436	8.0	392	4.6	379	4.7
History of CVD§	314	5.9	328	6.0	268	3.2	287	3.5
BMI, kg/m2, median [Q3-Q1]	29.2	[25.7–33.7]	29.2	[25.7–33.5]	27.5	[24.2–31.7]	27.5	[24.3–31.7]
Waist circumference, cm, mean (SD)	91.7	(14.0)	91.5	(13.7)	88.0	(13.7)	88.0	(13.8)
Systolic BP, mm Hg, mean (SD)	130.4	(17.5)	130.2	(17.6)	127.6	(17.6)	127.8	(17.5)
Diastolic BP, mm Hg, mean (SD)	76.5	(9.2)	76.5	(9.4)	75.6	(9.1)	75.8	(9.1)

CEE indicates conjugated equine estrogens; MPA, medroxyprogesterone acetate; SD, standard deviation; y, year; CVD, cardiovascular disease; VMS, vasomotor symptoms; Q3, 75^th percentile; Q1, 25^th percentile.

^*Numbers may not sum to column totals because of minimal missing data; missingness was ≤1.3% for any baseline characteristic within each arm. Percentages may not sum to 100 because of rounding.

^†Self-reported against six fixed categories; women who self-identified as “Other” were classified as “None of the above.”

^‡Required a 3-month washout prior to randomization.

^§Includes MI, stroke, and venous thromboembolism.

Overall HRs for CHD were 0.94; 95%CI: 0.78–1.14 for CEE-alone versus placebo and 1.19; 95%CI: 0.97–1.47 for CEE+MPA versus placebo (Figure 2).⁷ Biomarkers were assessed for 714 of 781 total cases (91.4%), compared to 359 or fewer in all prior reports.^15–18 Participants without prior treatment for hyperlipidemia) had HRs for CHD of 0.99; 95%CI: 0.79–1.24 in the CEE-alone trial and 1.29; 95%CI: 1.01–1.64 in the CEE+MPA trial compared to respective placebo. Among participants with no prior treatment for hyperlipidemia, the risks of CHD for CEE+MPA versus placebo increased with increasing levels of LDL cholesterol, non-HDL cholesterol, and LDL/HDL cholesterol ratios. Specifically, HRs for CHD in the CEE+MPA versus placebo groups for participants with LDL <130 mg/dl was 0.59; 95% CI: 0.31–1.10; for ≥190 mg/dl HR was 2.77; 95%CI: 1.42–5.40, P-trend=0.002; for non-HDL cholesterol <160 compared to ≥220 mg/dl HRs were 0.73; 95%CI: 0.42–1.28 and 2.15; 95%CI: 1.18–3.92, P-trend=0.007; for LDL/HDL ratio <2.5 compared to ≥4 HRs were 0.73 (95%CI: 0.39–1.37), and 1.76; 95%CI: 1.08–2.88, P-trend=0.008 (Figure 2). A similar but non-significant pattern of risk by LDL/HDL ratio were seen for CEE-alone versus placebo. Participants with HDL cholesterol levels ≥60 mg/dl compared to <50 mg/dl were at somewhat lower risk in the CEE-alone group (HRs 0.68; 95%CI: 0.35–1.30 and 1.24; 95%CI:0.93–1.66, P-trend=0.02).

Figure 2.

Overall effect of menopausal hormone therapy on coronary heart disease and select subgroups. Overall hazard ratios (HRs) from Cox models that included covariate adjustment for age (linear), race and ethnicity, prior menopausal hormone therapy (MHT) use, and history of cardiovascular disease, with baseline hazard function stratified by age group, Dietary Modification trial, current vasomotor symptoms (VMS), and history of myocardial infarction. Subgroup analyses included a treatment × subgroup interaction and additionally stratified by subgroup. Annualized rates were computed as total number of events divided by total follow-up time, y, expressed as a percentage. P-values correspond to a test for the main effect (i.e., subgroup label “Overall effect on CHD”), interaction between randomization group × history of treatment, or 1-df interaction (trend) test between randomization group × ordered cardiometabolic groups (assigned integer values) among women without a history of treatment (e.g., “LDL-C, mg/dL”). P-values were obtained from a pooled Wald test by combining regression coefficients and standard errors across imputed datasets using Rubin’s rules. *Self-reported high cholesterol requiring pills or current use of antihyperlipidemic medications. Women without such history were categorized by lipid concentration. †Self-reported ever taking pills for high blood pressure or current use of antihypertensive medications. Women without such history were categorized by measured blood pressure. ^‡Self-reported ever taking pills or insulin shots for diabetes or current use of antidiabetic medications. Women without such history were categorized by glucose concentration. ^§Meeting ≥3 of 5 Adult Treatment Panel III criteria: waist circumference ≥88 cm, triglycerides ≥150 mg/dL, high-density lipoprotein cholesterol (HDL)-C <50 mg/dL, systolic blood pressure (BP) ≥130 or diastolic BP ≥85 mm Hg, and fasting glucose ≥100 mg/dL. Women reporting treatment for hyperlipidemia, hypertension, or diabetes were separately categorized. CEE, conjugated equine estrogens; MPA, medroxyprogesterone acetate; PY, person-years; CHD, coronary heart disease; LDL-C, low-density lipoprotein cholesterol.

In contrast to the findings for cholesterol, levels of untreated blood pressure, glucose, triglycerides or presence of MetS did not modify the effects of CEE-alone or CEE+MPA versus placebo on CHD risks. As observed in the placebo groups, participants with a history of treated hyperlipidemia, hypertension, diabetes, or any component of MetS had higher baseline absolute rates of CHD than untreated participants, but randomization to CEE-alone or CEE+MPA versus placebo did not further increase their risk for CHD.

The patterns for effect modification by untreated LDL cholesterol levels within 10-year age groups (50–59, 60–69, and 70–79 years) were apparent and consistent with the overall results in the CEE+MPA trial, and though confidence intervals were wide, interaction tests for LDL cholesterol were significant in the age groups 50–59 and 70–79 years (Appendix 4, available online at http://links.lww.com/xxx). Patterns for HDL cholesterol within age groups in the CEE trial were consistent with the overall results but tests for interaction were not significant.

During the initial 2 years after randomization, the overall risk of CHD due to oral MHT was non-significantly greater in the CEE-alone trial (HR 1.14; 95%CI:0.77–1.68) but was increased significantly in the CEE+MPA trial (HR 1.50; 95%CI: 1.06–2.13) (Figure 3).^7–9,26 Participants with no prior history of treatment for hyperlipidemia randomized to CEE had an overall HR for CHD of 1.24; 95%CI: 0.77–1.99 and those randomized to CEE+MPA an overall HR for CHD of 1.59; 95%CI: 1.05–2.41 versus placebo. CHD risk was increased in participants with LDL cholesterol >190 mg/dl (HR 3.20; 95%CI: 1.17–8.77) or non-HDL cholesterol >220 mg/dl (HR 2.78; 95% CI: 1.16–6.63) for CEE+MPA versus placebo, but tests for trend were not significant. Results for the effects of CEE-alone versus placebo by lipid levels were not significant. Untreated blood pressure levels, glucose levels, triglycerides or MetS did not modify the effect of oral MHT versus placebo on CHD in either trial.

Figure 3.

Overall effect of hormone therapy on coronary heart disease (CHD) and select subgroups at two years. Overall hazard ratios (HRs), annualized rates, and P-values described in Figure 2, except with follow-up administratively censored two years after randomization. *Self-reported high cholesterol requiring pills or current use of antihyperlipidemic medications. Women without such history were categorized by lipid concentration. †Self-reported ever taking pills for high blood pressure or current use of antihypertensive medications. Women without such history were categorized by measured blood pressure. ^‡Self-reported ever taking pills or insulin shots for diabetes or current use of antidiabetic medications. Women without such history were categorized by glucose concentration. ^§Meeting ≥3 of 5 Adult Treatment Panel III criteria: waist circumference ≥88 cm, triglycerides ≥150 mg/dL, high-density lipoprotein cholesterol (HDL-C) <50 mg/dL, systolic blood pressure (BP) ≥130 or diastolic BP ≥85 mm Hg, and fasting glucose ≥100 mg/dL. Women reporting treatment for hyperlipidemia, hypertension, or diabetes were separately categorized. CEE, conjugated equine estrogens; MPA, medroxyprogesterone acetate; MHT, menopausal hormone therapy; PY, person-years; LDL-C, low-density lipoprotein cholesterol.

For both trials, the increased risks for stroke and VTE, as well as neutral effects on all-cause mortality, were similar to those previously reported, and did not differ by CVD risk factor treatment status or by level of untreated risk factors (Figure 4, Appendix 5, Appendix 6, available online at http://links.lww.com/xxx).^7,10–13,27 However, all-cause mortality was increased for participants with a history of treated hypertension randomized to CEE-alone (HR 1.28; 95%CI: 1.04–1.58, P-interaction 0.004); of the 25 excess deaths/10,000/year 7 were CVD, 4 were cancer, and 14 were non-CVD non-cancer deaths.

Figure 4.

Overall effect of hormone therapy on stroke and select subgroups. Overall hazard ratios (HRs), annualized rates, and P-values described in Figure 2, except with Cox model stratified on history of stroke instead of history of myocardial infarction. *Self-reported high cholesterol requiring pills or current use of antihyperlipidemic medications. Women without such history were categorized by lipid concentration. †Self-reported ever taking pills for high blood pressure or current use of antihypertensive medications. Women without such history were categorized by measured blood pressure. ^‡Self-reported ever taking pills or insulin shots for diabetes or current use of antidiabetic medications. Women without such history were categorized by glucose concentration. ^§Meeting ≥3 of 5 Adult Treatment Panel III criteria: waist circumference ≥88 cm, triglycerides ≥150 mg/dL, high-density lipoprotein cholesterol (HDL-C) <50 mg/dL, systolic blood pressure (BP) ≥130 or diastolic BP ≥85 mm Hg, and fasting glucose ≥100 mg/dL. Women reporting treatment for hyperlipidemia, hypertension, or diabetes were separately categorized. CEE, conjugated equine estrogens; MPA, medroxyprogesterone acetate; MHT, menopausal hormone therapy; PY, person-years; LDL-C, low-density lipoprotein cholesterol.

DISCUSSION

This report confirms previous findings of effect modification by baseline lipids but does not confirm effect modification by MetS status. It extends previous findings by separating out participants with a history of treatment for CVD risk factors from untreated participants to obtain estimates of effect modification not confounded by risk factor treatment effects. In participants without a history of treated hyperlipidemia the risk of CHD due to CEE-alone or CEE+MPA varied by lipid level. Participants not treated for hyperlipidemia with the healthier lipid profiles (i.e. lowest LDL cholesterol, non-HDL cholesterol levels, LDL/HDL cholesterol ratios, or highest HDL cholesterol levels) did not appear to be at increased risk for CHD in either trial. However, randomization to CEE+MPA versus placebo significantly increased risk in participants with LDL cholesterol ≥190 mg/dl, non-HDL cholesterol ≥220 mg/dl, or LDL/HDL ratio ≥3. CEE showed less effect modification by lipid levels, but patterns were like those found in the CEE+MPA trial, especially for HDL cholesterol and LDL/HDL cholesterol ratio. Effect modification by lipid levels were generally consistent within age groups; thus, the overall effects can be assumed to apply to all age strata. Untreated blood pressure, blood glucose, and triglyceride levels did not influence the effect of oral MHT on CHD. The effects of oral CEE-alone or CEE+MPA on stroke, all-cause mortality, and VTE were not influenced by untreated lipids or other CVD risk factors studied.

Higher absolute baseline risks for CHD in participants with a history of treated hyperlipidemia, hypertension, diabetes, or one of the components of the MetS likely reflect that those with the highest levels or longest duration of a CVD risk factor, and thus at highest risk, are more likely to receive treatment for that risk factor, but such treatment does not result in a return to low risk status (Appendix 3, available online at http://links.lww.com/xxx). Oral MHT did not further increase risk in treated participants, suggesting that CVD risk factor treatment mitigated potential adverse effects of MHT, however it is not possible to separate out the effects of being under medical care from those directly due to the treatment. The results are consistent with previous findings from WHI in which participants treated with statins tended to have lower risks for CHD due to oral MHT than untreated participants, with a significant difference in favor of statins at 3.5 years post-randomization in participants randomized to CEE-alone compared to placebo.²⁶

The increased CHD risk due to oral MHT during the first 2 years after randomization also appeared to be modified by lipid levels, with significantly increased risk for CEE+MPA versus placebo in participants with LDL cholesterol ≥190 mg/dl or non-HDL cholesterol ≥220 mg/dl, and somewhat increased risk for CEE-alone versus placebo in participants with HDL cholesterol <50 mg/dl. The findings for effect modification during the first years of oral MHT are of interest, because CHD risk depends on duration after initiation in addition to age or years since menopause.^7–9,26 The time course has been well studied for CEE+MPA but may also apply to a lesser extent for CEE-alone which has a less adverse effect on CHD risk.^7–9,26 In the HERS secondary CHD prevention trial of CEE+MPA the HR=1.52 in the first year attenuating or reversing in later years to end with an overall HR=0.99⁴; there are similar findings of early harm in observational studies of MHT initiated in women with prior MI.^28,29 Regarding primary prevention, in the WHI trial of CEE+MPA the HR=1.50 in the first 2 years attenuating to overall HR=1.19 in the current analysis; in the Nurses’ Health Study (NHS) observational study when analyzed as a series of clinical trials CEE+MPA initiators had HR=1.42 in first 2 years attenuating or reversing to overall HR=0.96.³⁰ When the Women’s International Study of long Duration Oestrogen after Menopause (WISDOM) was stopped after 1 year, there was a significant increase in the number of major cardiovascular events (7 vs. 0, P=0.016) for CEE+MPA versus placebo.³¹ While clinical trials capture the early events, many conventionally-analyzed observational studies do not, and this may contribute to the more favorable findings from observational studies compared to clinical trials.³² Similarities in the time course of CHD risk for secondary and primary prevention can be attributed to the high prevalence of subclinical atherosclerosis in the general population. Though lower than in men, subclinical coronary atherosclerosis is present in 18.8, 29.3, and 40.1% of women aged 50–54, 55–59, and 60–64 years, respectively.³³ Atheroma burden increases in older women, and hence they have a greater absolute risk of CHD. The early increase in risk of CHD is likely due to destabilization of existing plaques by oral MHT potentially though increased tissue expression and activation of matrix metalloproteinases (MMPs).^34,35 The pro-thrombotic effect of oral MHT would facilitate subsequent thrombosis and vessel occlusion. Prior work in the WHI trials showed that higher baseline levels of MMP-9 were associated with CHD in the first 2 years, and that randomization to MHT markedly increased circulating MMP-9 levels and prothrombotic factors at 1 year.¹⁵ On the other hand, oral MHT has favorable effects on blood lipids, reducing LDL cholesterol and lipoprotein(a) levels and increasing HDL cholesterol levels at year 1 and over longer term follow-up.^15,36Destabilization of existing plaques may predominate initially leading to an early increase in coronary events but over time favorable lipid effects may stabilize remaining plaques or even reduce new atheroma formation leading to an attenuation of risk. CEE+MPA may be more adverse than CEE-alone because it is associated with greater increase in MMP-9 levels as well as being less effective at increasing HDL cholesterol levels.^15,36

The effects of other forms of MHT on biomarkers have not been extensively studied. Small clinical trials indicate that oral estradiol and CEE have similar effects on blood lipids and markers of hemostasis and inflammation.^37,38 Transdermal estradiol has little effect on blood lipids, hemostatic factors, or metalloproteinases.^37.38 Micronized progesterone blunts the estrogen-induced increase in HDL cholesterol to a lesser extent than MPA.³⁹ However, there is no large long-term clinical outcomes trial directly assessing the safety of oral or transdermal estradiol with or without progesterone.

Potential explanations for the increased risk for all-cause mortality in participants treated for hypertension in the CEE-alone trial include interaction of off-target effects of antihypertensive treatment with CEE; however it is more likely to be a chance finding since similar effects were not seen in the CEE+MPA trial and there was no cause-specific pattern to the deaths.

Limitations in this study include that subgroup analyses are exploratory and some subgroups had relatively sparse data with wide confidence intervals. Due to the large number of interactions studied, some results may have been significant by chance. Only oral CEE and MPA were studied and these findings may not apply to other formulations and routes of delivery.

Clinical implications include that measurement and treatment of dyslipidemia, high blood pressure, blood glucose, and components of MetS might mitigate or remove additional cardiovascular risks due to oral MHT. Lipid management may be uniquely important because of multiplicative interactions of lipids with CHD risk due to oral MHT, whereas non-lipid risk factors appear to be additive to any risk from MHT. While it is possible that early statin therapy might be beneficial, aspirin does not appear to influence the risk of CHD due to CEE-alone or CEE+MPA.²⁶ Measurement of blood lipids could assist with tailoring MHT to the women most likely to benefit. Further research is needed into methods of blocking the initial destabilization, rupture and thrombosis of vulnerable plaques. A large clinical outcomes trial of oral and transdermal estradiol with and without progesterone is needed to support emerging changes in practice.

Data Sharing Statement:

• Will individual participant data be available (including data dictionaries)? YES

• What data in particular will be shared?: ALL DATA RELATED TO THE TRIALS AND THIS SECONDARY ANALYSIS

• What other documents will be available?: STUDY PROTOCOL AND STATISTICAL ANALYSIS PLAN WILL BE AVAILABLE

• When will data be available (start and end dates)?: UPON PUBLICATION AND INDEFINITELY

• By what access criteria data will be shared (including with whom, for what types of analyses, and by what mechanism)?: DATA WILL BE MADE AVAILBLE TO RESEARCHERS AFTER APPROVAL OF STUDY PROPOSAL. DETAILS CAN BE FOUND AT WWW.WHI.ORG