eFigure 1. Stratified and Sensitivity Analyses for Venous Thrombosis
eFigure 2. Stratified and Sensitivity Analyses for Myocardial Infarction
eFigure 3. Stratified and Sensitivity Analyses for Ischemic Stroke
eFigure 4. Stratified and Sensitivity Analyses for Measures of APC Resistance
Smith NL, Blondon M, Wiggins KL, Harrington LB, van Hylckama Vlieg A, Floyd JS, Hwang M, Bis JC, McKnight B, Rice KM, Lumley T, Rosendaal FR, Heckbert SR, Psaty BM. Lower Risk of Cardiovascular Events in Postmenopausal Women Taking Oral Estradiol Compared With Oral Conjugated Equine Estrogens. JAMA Intern Med. 2014;174(1):25-34. doi:10.1001/jamainternmed.2013.11074
Little is known about the comparative cardiovascular safety of oral hormone therapy products, which impedes women from making informed safety decisions about hormone therapy to treat menopausal symptoms.
To compare the relative clinical cardiovascular safety of 2 commonly used oral estrogen drugs—conjugated equine estrogens (CEEs) and estradiol.
Design, Setting, and Participants
Population-based, case-control study from January 1, 2003, to December 31, 2009, comparing cardiovascular event risk associated with current CEEs and estradiol use in a large health maintenance organization in which the preferred formulary estrogen changed from CEEs to estradiol during the course of data collection. Participants were 384 postmenopausal women aged 30 to 79 years using oral hormone therapy.
Main Outcomes and Measures
Incident venous thrombosis was the primary clinical outcome, and incident myocardial infarction and ischemic stroke were secondary outcomes. As validation, an intermediate clotting phenotype, the endogenous thrombin potential–based normalized activated protein C sensitivity ratio, was measured in plasma of controls.
We studied 68 venous thrombosis, 67 myocardial infarction, and 48 ischemic stroke cases, with 201 matched controls; all participants were current users of oral CEEs or estradiol. In adjusted analyses, current oral CEEs use compared with current oral estradiol use was associated with an increased venous thrombosis risk (odds ratio, 2.08; 95% CI, 1.02-4.27; P = .045) and an increased myocardial infarction risk that did not reach statistical significance (odds ratio, 1.87; 95% CI, 0.91-3.84; P = .09) and was not associated with ischemic stroke risk (odds ratio, 1.13; 95% CI, 0.55-2.31; P = .74). Among 140 controls, CEEs users compared with estradiol users had higher endogenous thrombin potential–based normalized activated protein C sensitivity ratios (P < .001), indicating a stronger clotting propensity.
Conclusions and Relevance
In an observational study of oral hormone therapy users, CEEs use was associated with a higher risk of incident venous thrombosis and possibly myocardial infarction than estradiol use. This risk differential was supported by biologic data. These findings need replication and suggest that various oral estrogen drugs may be associated with different levels of cardiovascular risk.
The therapeutic effectiveness of oral estrogens to treat menopausal symptoms is well established and is similar for the most commonly used estrogen products of conjugated equine estrogens (CEEs) and estradiol.1,2 However, little is known about the relative safety of these oral hormone therapy (HT) products regarding cardiovascular events.
In a population-based, case-control study among postmenopausal women using oral HT, we tested the hypothesis that oral CEEs are more prothrombotic than oral estradiol. Extrapolating findings from previous research on CEEs risk,3,4 we hypothesized that current use of oral CEEs would be associated with a greater risk of a first venous thrombosis (VT) than current estradiol use. Furthermore, we evaluated risk for a first myocardial infarction (MI) and first ischemic stroke and hypothesized associations in the same direction as VT. To provide biologic validity to our hypotheses, we measured differences in biologic thrombotic propensity among controls using CEEs or estradiol.5,6
The setting was the Heart and Vascular Health Study, a case-control study of incident cardiovascular events and frequency-matched controls in men and women aged 30 to 79 years who are members of Group Health Cooperative (GHC), a large health maintenance organization in Washington State.3,7- 10 Cases included VT, MI, stroke, and atrial fibrillation events, and the index date was the date of their incident event or the date of death for out-of-hospital fatal events. Case identification began on January 1, 1989, and complete data through December 31, 2009, were available for this analysis. The control group was frequency matched 3:1 to the largest case group (MI) by sex, age within decade, treated hypertension status, and calendar year of identification. Their index date was a randomly chosen date within the calendar year from which they were selected as a control. The study was approved by the GHC Human Subjects Review Committee and written consent was provided by those who gave a blood sample.
For this comparative safety study of estrogen type, we restricted participants to postmenopausal women using oral HT from January 1, 2003, forward. This date was chosen because HT use changed dramatically in the GHC population after the June 2002 publication of the first Women’s Health Initiative trial results.11,12
Cases experienced an incident MI, ischemic stroke, or VT (deep vein thrombosis or pulmonary embolism) and had no history of VT, MI, or ischemic stroke events. All were using oral CEEs or estradiol at the time of their event and were not using anticoagulants. Events (VT, MI, and ischemic stroke) were identified using hospital discharge International Classification of Diseases, Ninth Revision (ICD-9) diagnosis codes from GHC and non-GHC facilities. In addition, VT events were identified using ICD-9 codes assigned at urgent care clinic visits, and GHC pharmacy records were reviewed to identify women who were dispensed a prescription for a low-molecular-weight heparin for outpatient deep vein thrombosis treatment.
Trained abstractors reviewed the medical records of all potential cases to verify the diagnosis of VT, MI, or ischemic stroke and to determine how the diagnosis was made. Events criteria were adapted from the Leiden Thrombophilia Study13 for VT and from the Cardiovascular Health Study14,15 for MI and ischemic stroke. Venous thrombosis events were classified as study eligible if they were diagnosed by an imaging modality, if the event was a fatal pulmonary embolism based on ICD-9 or ICD-10 diagnostic information, or if the event was diagnosed by a physician based on the presence of symptoms in the absence of imaging. Ninety-three percent of eligible VT cases had positive diagnostic imaging test results. For MI, verification was based on symptoms, electrocardiogram findings, cardiac enzyme levels, and physician diagnosis and treatment. For ischemic stroke, verification required the rapid onset of a neurologic deficit that persisted at least 24 hours or evidence of infarction on brain imaging studies,7,9 which was present in 98% of the cases.
Controls included in this analysis were postmenopausal and had no history of VT, MI, or ischemic stroke. All were using oral CEEs or estradiol at the time of their index date and were not using anticoagulants.
Use of HT was ascertained from the GHC computerized pharmacy database. More than 95% of GHC members in the age range of our participants fill all or almost all (90%-100%) of their prescriptions through GHC pharmacies.16 The detailed pharmacy data include for each GHC member identification number the drug name, date the prescription was filled, medication strength and quantity of medication prescribed, and dosing instructions or the number of days the supply of medication was intended to last.
The GHC formulary contains preferred agents within each drug class. From January 1, 2003, through January 31, 2005, the preferred estrogen was oral CEEs. On February 1, 2005, to save on drug costs, GHC changed the preferred agent to oral estradiol; it has remained the preferred agent. Switching estrogens was encouraged by GHC, but women were allowed to continue taking the more expensive, nonpreferred drug if they chose, but it would require an additional copayment. In this study, 87% of women were using the preferred estrogen at their index date. Other oral estrogens, such as esterified estrogen and transdermal estrogens, were rarely used during 2003 to 2009, and women using them were excluded from the analysis.
A woman was considered a current user of oral HT if she received enough medication with her final prescription before the index date to last until her index date given at least 80% compliance with treatment. Daily oral estrogen dose was calculated from computerized pharmacy data using pill strength and dosing instructions or days’ supply. Recency of HT initiation was defined as the number of days between the index date and the date of the first prescription fill for oral HT. Concomitant use of a progestogen was calculated similarly using an 80% compliance adjustment. Among progestogen users, 96% were using medroxyprogesterone acetate.
Demographic and health status information was obtained by review of the entire GHC ambulatory medical record up to the index date. The number of GHC visits in the year before the index date was recorded, as well as birth date, height, weight, and race/ethnicity. Menopausal status at the index date was based on information collected from the GHC medical record for cases and controls. A woman was considered postmenopausal if she had experienced the onset of menopausal symptoms or a cessation of ovarian function that occurred naturally or through a bilateral oophorectomy. If not explicitly stated in the medical record, postmenopausal status was assigned to women at 55 years and older. We collected information on risk factors for venous and arterial thrombotic events, including information on smoking, general health, hypertension, diabetes mellitus, cholesterol level, history of cardiovascular conditions and procedures, cancer (all cancers except nonmelanoma skin cancers), and hospitalizations or inpatient surgical procedures in the last 30 days. Additional data were collected from surviving individuals who were invited to participate in a telephone interview. Missing data were uncommon, and no more than 5% of the data for any variable was imputed using multiple imputation.
Individuals who participated in the interview were invited to donate a venous blood sample. The first 4.5 mL of blood was collected into a tube of 3.2% sodium citrate. Samples were centrifuged at 4°C for 10 minutes at 1300g and stored at −70°C.
Thrombotic potential was assessed using the endogenous thrombin potential (ETP)–based normalized activated protein C (APC) sensitivity ratio (nAPCsr), which quantifies resistance to APC and has been shown to be a strong predictor of VT in the setting of estrogen treatment.5,17 Citrated plasma samples collected from controls were shipped on dry ice from Seattle, Washington, to Leiden University Medical Center, Leiden, the Netherlands, for analysis. Plasma, which was thawed once before, was used to measure ETP-based APC resistance directly using a fluorogenic assay (Thrombinoscope TM; Synapse BV), which measures thrombin generation. The ETP was calculated from the area under the thrombin generation curve, and the nAPCsr was the normalized ratio of the amount of thrombin generated in the absence of APC to the quantity generated in the presence of APC.18 Higher values for ETP-nAPCsr indicate a more prothrombotic profile.
Multiple logistic regression analysis was used to assess the association of the type of currently used estrogen and occurrence of VT, MI, and ischemic stroke, separately. Adjustment variables included design and a priori potential confounding variables. For VT analyses, we included the following variables: age, current smoking status, current statin or progestogen use, and estrogen daily dose using categories of low (<0.625 mg for CEEs and <1.0 mg for estradiol) or modal or high (≥0.625 mg for estradiol and ≥1.0 mg for estradiol), as well as race/ethnicity (white vs other), treated diabetes mellitus or treated hypertension status, the number of GHC visits in the year before the index date, body mass index (calculated as weight in kilograms divided by height in meters squared), cancer history (cancer diagnosed within 5 years before and 3 months after the index date), and prevalent cardiovascular disease (history of angina, claudication, peripheral vascular disease, or cardiovascular or carotid revascularization). For MI and ischemic stroke analyses, we added total cholesterol level and systolic blood pressure variables and removed the cancer variable. To take advantage of the quasiexperiment provided by the formulary change in the standard estrogen preparation, we included the log sampling ratio for the reference year as an offset in the logistic model rather than using regression adjustment for the reference year; this provided odds ratio (OR) estimates for the marginal association between case status and estrogen formulation standardized to the reference year distribution of cases. Results are presented throughout as estimated ORs (95% CIs), calculated using robust SEs.
Stratified analyses fit separate models in subgroups according to progestogen use, age (<60 vs ≥60 years), recency of HT initiation (≤3 vs >3 years), and estrogen daily dose (low vs modal or high) and tested differences in the ORs across strata using multiplicative models. Sensitivity analyses of the primary findings were conducted assuming 100% compliance with HT instead of 80% and restricting analyses to include only those using the preferred agent on the GHC formulary at their index date.
We assessed the thrombotic potential of the 2 types of estrogen drugs. Among a subset of 140 controls who agreed to phlebotomy, we compared levels of log-transformed ETP-based nAPCsr between CEEs and estradiol users using multiple linear regression analysis with robust SEs and adjusting for potential confounders. Current hormone type and daily dose were based on use at phlebotomy; adjustment variable information was obtained at the index date. The differences in hemostatic measure levels are reported as a ratio of CEEs to estradiol of their adjusted geometric means. Secondary analyses paralleled those used with the clinical outcomes, as described above.
From January 1, 2003, through December 31, 2009, there were 68, 67, and 48 postmenopausal women who experienced an incident VT, MI, or ischemic stroke event, respectively, and 201 controls, all of whom were current users of oral CEEs or estradiol. The participants were mostly of white race/ethnicity, with a mean age ranging from 63.2 to 67.6 years (Table 1). Cases were more likely than controls to have arterial and VT risk factors. Among controls (Table 2), those using oral estradiol had a slightly larger risk factor burden than those using CEEs, but no major differences were observed in age, body mass index, self-reported general health status, or recency of HT initiation.
In analyses adjusting for design and potential confounding variables (Table 3), we observed a greater risk of VT associated with current oral CEEs use than with current oral estradiol use (OR, 2.08; 95% CI, 1.02-4.27; P = .045). For the outcome of MI, current oral CEEs use was also associated with an increase in MI risk, although less clearly (OR, 1.87; 95% CI, 0.91-3.84; P = .09). No increased risk was observed for ischemic stroke (OR, 1.13; 95% CI, 0.55-2.31; P = .74).
When the population was stratified by age, estrogen daily dose, recency of HT initiation, and concomitant progestogen use, we found no statistical evidence of interaction (interaction P > .09 for all) (eFigures 1, 2, and 3 in the Supplement). In subset sensitivity analyses that assumed 100% compliance with HT use and that restricted use to those using the preferred estrogen, relative risk estimates were generally similar but less precise.
The median ETP-based nAPCsr levels were 1.96 (interquartile range, 1.40-2.78) among 94 controls using oral estradiol and 2.68 (interquartile range, 2.17-7.18) among 46 controls using oral CEEs. The geometric mean nAPCsr level among CEEs users was 68% greater than that of estradiol users (ratio of CEEs to estradiol, 1.68; 95% CI, 1.24-2.28; P < .001) after adjusting for age, race/ethnicity, current smoking, body mass index, factor V Leiden variant, self-reported general health, cancer, and treated diabetes mellitus status, as well current statin use, progestogen use, and estrogen daily dose. In stratified analyses, there was no strong statistical evidence of differences by subgroup (eFigure 4 in the Supplement).
In this contemporary observational study of women using HT, we found that current oral CEEs use was associated with a greater risk of an incident VT than current oral estradiol use. We also found that among controls women using oral CEEs had a greater resistance to APC than women using oral estradiol, providing biologic support for the increased risk seen in the case-control analysis. For arterial end points, MI relative risk estimates were similar to estimates for VT but did not reach statistical significance. For ischemic stroke, our data did not suggest relative risk differences by estrogen type. Although no statistical differences were observed in subgroups, the relative risks of VT and MI were only slightly elevated for low-dose CEEs users compared with low-dose estradiol users. For VT, there was also a suggestion that relative risks were elevated for those using progestogen and for recent HT initiators compared with non–progestogen users and with longer-term users, findings that parallel those of the Women’s Health Initiative,11,19 which compared CEEs with placebo.
Cardiovascular risk associated with HT has been widely discussed, especially as it concerns timing of HT initiation relative to menopause, but far less attention has been directed toward the comparative safety of these products in women who initiate therapy to treat the vasomotor symptoms of menopause.20 To our knowledge, no trial is under way that includes clinical cardiovascular end points. Among the published observational studies21- 27 of HT that include clinical cardiovascular end points, data are limited and have focused primarily on the route of HT delivery, comparing oral with transdermal formulations. The comparative safety of oral esterified estrogen compared with oral CEEs has been previously described, and it was found that oral CEEs use compared with oral esterified estrogen use was associated with an increased risk of VT and possibly MI and ischemic stroke; findings were validated in controls, where oral CEEs users had greater resistance to APC than oral esterified estrogen users.3,4,28 Furthermore, recent data suggest that estrogen and progestogen drug type may be differentially associated with cardiovascular event risk, although primary analyses compared HT use with nonuse.26,27
In this study, we provide information on the relative cardiovascular safety of 2 commonly used oral estrogens but do not give data about absolute risks or risks relative to women not using HT. Women who seek HT for menopausal symptom treatment are different from those who do not, and accounting for these differences in observational studies has been problematic.29
In this observational comparative safety study, the type of estrogen a woman received was not determined by randomization but rather by when the woman received estrogen. This quasiexperimental design, where differences in treatment approaches are largely systemwide and not related to patient-specific factors, overcomes some biases associated with nonrandom treatment allocation. That our study compares only HT users (a requirement for a comparative safety study of HT approaches) provides additional strength because we avoid issues of confounding by indication because all women were using HT.
Our case-control design captured users of oral HT at the time of sampling, regardless of their duration of HT use, and does not address the potential for selection bias because of differential initiation or discontinuation of therapy. Therefore, our risk estimates are in relation to active use of oral HT and do not directly address the relative risk of initiating HT. However, the results of our subgroup analyses suggest that we were able to capture (to some extent) VT risk differences by the time since HT initiation. Given that our study design was more likely to capture initiators of estradiol than CEEs, the CEEs initiators were underrepresented in our study, and a potential bias may have been to underestimate CEEs risk relative to estradiol risk. Furthermore, residual confounding could be present if factors associated with changes in HT use and cardiovascular diagnosis throughout the study period were not fully adjusted for. A new-user study design among women initiating HT who are followed up for future thrombotic events would provide such information.
There are several other strengths of the study. These include the population-based design, the verification of thrombotic outcome events, the use of GHC pharmacy data to capture detailed HT exposure before the index date, and the inclusion of both biologic and clinical phenotypes to assess thrombotic risk associations.
Pharmaceutical estrogens are molecularly distinct. Conjugated equine estrogens are manufactured from the urine of pregnant mares and contain 10 known biologically active estrogen compounds, as well as others that have yet to be described.30,31 The primary compounds are estrone and equilin sulfate.31,32 Estradiol is a “natural” or “bioequivalent” estrogen, containing only synthetic estradiol-17β. How estrogenic compounds modulate thrombotic risk remains poorly understood, and our data suggest that risk potential may vary by the type or amount of estrogen-derived molecules within various marketed estrogens.3,4,28,32
The findings of this comparative safety investigation need replication. If confirmed, the results would provide valuable information to women and their health care professionals when making safety decisions regarding available HT options for menopausal symptom management. The number of avoidable venous and possibly cardiac thrombotic events may be lowered by selecting lower-risk medications.
In summary, we found in an observational study of HT use that women using oral CEEs were at an increased risk of incident VT compared with women using oral estradiol. This finding was supported by biologic data on APC resistance. Our data also suggested an increase in the risk of MI but not ischemic stroke for CEEs use compared with estradiol use. These results need replication but suggest that oral estrogen drugs have different levels of cardiovascular risk.
Accepted for Publication: August 2, 2013.
Corresponding Author: Nicholas L. Smith, PhD, Department of Epidemiology, University of Washington, 1730 Minor Ave, Ste 1360, Seattle, WA 98101 (firstname.lastname@example.org).
Published Online: September 30, 2013. doi:10.1001/jamainternmed.2013.11074.
Author Contributions: Dr Smith had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Smith, Blondon, Lumley, Rosendaal, Heckbert, Psaty.
Acquisition of data: Smith, Heckbert, Psaty.
Analysis and interpretation of data: Smith, Blondon, Wiggins, Harrington, van Hylckama Vlieg, Floyd, Bis, McKnight, Rice, Lumley, Rosendaal, Heckbert, Psaty.
Drafting of the manuscript: Smith.
Critical revision of the manuscript for important intellectual content: Blondon, Wiggins, Harrington, van Hylckama Vlieg, Floyd, Bis, McKnight, Rice, Lumley, Rosendaal, Heckbert, Psaty.
Statistical analysis: Blondon, Wiggins, McKnight, Rice, Lumley.
Obtained funding: Smith, Heckbert, Psaty.
Administrative, technical, or material support: Smith, Wiggins, Bis, Rosendaal, Psaty.
Study supervision: Smith, Rosendaal, Psaty.
Conflict of Interest Disclosures: Dr Blondon is a recipient of a fellowship for prospective researchers from the Swiss National Science Foundation. Dr Rosendaal is listed on several patents of prothrombotic gene variants, including factor V Leiden, prothrombin 20210A, fibrinogen gamma 10034T, and several FXI variants; royalties received from licenses from these patents are paid to his institution. Dr Psaty serves on a data safety and monitoring board for a clinical trial of a device funded by Zoll LifeCor and on a steering committee for the Yale Open Data Access Project funded by Medtronic, Inc.
Funding/Support: The Heart and Vascular Health Study is supported by grants HL43201 (Psaty), HL60739 (Psaty), HL68986 (Heckbert), HL73410 (Smith), HL74745 (Psaty), HL85251 (Psaty), and HL95080 (Smith) from the National Heart, Lung, and Blood Institute.
Role of the Sponsor: The National Heart, Lung, and Blood Institute had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Correction: This article was corrected on August 6, 2014, to fix the Funding/Support section.