Methods
The WHI-OS is a large multicenter prospective cohort study that was conducted at 40 US sites. The details of the scientific rationale, eligibility criteria, and design of the WHI-OS have been previously published. Briefly, 93,676 postmenopausal women with or without intact uterus, aged 50 to 79 years, were recruited between September 1994 and December 1998, with clinic visits at baseline and 3 years. Annual follow-up by mailed self-administered questionnaires included detailed assessments of HT medications and collection of information on medical risk factors, lifestyle risk factors, and incident clinical events. CVD events were confirmed by medical record review. The present analyses include follow-up through August 14, 2009. Data were uniformly collected from participants by trained study staff according to a standardized institutional review board-approved protocol. All participants provided a written informed consent form for this research study at the time of enrollment.
As in previously published WHI research, major CHD was defined as nonfatal clinical myocardial infarction or death due to CHD. CVD and mortality outcomes were confirmed by physician adjudicators. Stroke was defined as the rapid onset of a neurologic deficit lasting more than 24 hours, supported by imaging studies. Total CVD included major CHD, stroke, and CVD mortality. Venous thromboembolism was not included because of the absence of medical record confirmation of this outcome in the WHI-OS. HT oral CEE doses were defined as follows: low-dose CEE, less than 0.625 mg; conventional-dose CEE dose, 0.625 mg; high-dose CEE, more than 0.625 mg. Oral estrogen formulation categories included oral estradiol and oral CEE. Oral estrogen plus progestogen (E + P) users included the formulations of both oral CEE and oral estradiol with a progestin or progesterone. Transdermal estrogen categorization included all dose formulations, as well as the use of concomitant oral progestin or progesterone among women with intact uterus.
Statistical Analysis
Cox proportional hazards models were used to calculate hazard ratios (HRs) and 95% CIs for different doses, routes of delivery, and formulations of HT as a time-varying exposure in relation to CVD outcomes, compared directly with conventional-dose oral CEE. Time to each CVD outcome was computed from the date of enrollment to the date of first CVD outcome event, and "censored" was defined as excluded from further follow-up by date of last study follow-up or August 14, 2009, whichever occurred first. Follow-up data on HT were collected from annual study questionnaires, with a mean follow-up time of 10.4 years. For each follow-up year, the type of hormone used was categorized as nonuser, estradiol, transdermal, oral low-dose CEE, oral conventional-dose CEE, conventional-dose CEE alone, E + P, and other. A variable for separate progestin or progesterone use was also created and adjusted for the analysis. All analyses were stratified by baseline 5-year age intervals and history of CVD, and adjusted for age (linear), race/ethnicity (white, black, Hispanic, Asian/Pacific Islander, American Indian/Alaskan Native, other), and smoking (never, former, current). Variables considered as potential confounders or effect modifiers were included as covariates in the model: body mass index (BMI) categories (<25, 25 to <30, ≥30 kg/m), BMI (linear), quartiles of total recreational physical activity, hypertension (never, untreated, treated), treated diabetes (no, yes), high cholesterol requiring medication or current use of lipid-lowering medication (no, yes), hysterectomy (no, yes), oophorectomy (no, partial, bilateral), educational attainment, and household income. Tests of proportional hazards assumption were conducted by testing the interaction terms of HT exposure, separately by dose and formulation, with time to the event or censoring for the five outcomes (major CHD, stroke, total CVD, CVD mortality, and all-cause mortality). All analyses were conducted using SAS version 9.2 (SAS Institute Inc, Cary, NC). All P values were two-sided tests, and values less than 0.05 were considered statistically significant.