Discussion
Our results indicate that CVD risk did not differ appreciably among women using different formulations, doses, and routes of administration of estrogen in comparison with conventional-dose oral CEE. Although similar rates were observed for most outcomes, oral estradiol may be associated with a lower risk of stroke, and transdermal HT and low-dose oral CEE may be linked to a lower risk of CHD, compared with conventional-dose oral CEE. The possibility that alternative formulations, doses, and routes of delivery may pose a lower risk of stroke and CHD than conventional-dose oral CEE is an important hypothesis that warrants confirmation in additional studies. The overall absolute risk of CVD and adverse events in younger women was much lower as compared with older women.
Our study is one of the first studies to provide head-to-head comparisons between the transdermal route of delivery and conventional-dose oral CEE on CVD outcomes. Although we found no statistically significant differences in CVD outcomes between these regimens, more favorable CHD findings with transdermal therapy were suggested. Additional research is needed to confirm these findings in view of the limited statistical power for these analyses. To date, large-scale randomized controlled trials of different HT formulations and risks of CVD events have not been conducted.
Our data add to emerging evidence suggesting that transdermal estrogen delivery may have advantages in minimizing the risk of CVD events associated with HT. A recent observational study of more than 80,000 women reported that oral, but not transdermal, HT carried an increased risk of stroke.
Transdermal treatments avoid the first-pass hepatic metabolism associated with an increase in thrombosis, although risks may be dose-dependent. The UK General Practice Research database found that the risk of stroke did not significantly increase with low-dose transdermal estradiol (dose formulation of 0.05 mg or lower), in contrast to an increased risk with higher transdermal and oral doses. We did not have sufficient power to stratify transdermal regimens by dose in the current study.
The timing of HT initiation in relation to proximity to menopause has been recognized as a potentially important predictor of CVD outcomes. This "timing hypothesis" has been bolstered by observational studies and by some of the recent analyses from the WHI-CT showing that women who initiate HT closer to the menopausal transition tend to have more favorable CHD outcomes with HT than those initiating HT late in menopause. Although we found similar HRs in younger and older women, our statistical power to address these issues was limited. Two recent clinical trials addressed the timing of HT initiation in relation to atherosclerosis progression, but they were not large enough to assess clinical CVD events.
The strengths of our study include the large, diverse, and well-characterized study population with a broad distribution of HT doses and formulations. In addition, all CVD events and deaths were physician-adjudicated. Information was available on a large number of potential confounding factors, including traditional CVD risk factors, measured BMI, physical activity, other lifestyle factors, and socioeconomic status (educational attainment and household income), which, in addition to age, were included as covariates in the model.
The limitations of our study include the observational design that precludes causal inference. In view of the observational study design, we cannot exclude the possibility that confounding and selection factors related to the choice of HT formulation may have contributed to the CVD risk reductions. A key difference between the WHI-OS and the WHI-CT, however, is that many of the women in the WHI-OS began HT closer to the onset of menopause, which may be less likely to precipitate CVD events than HT initiated in late menopause. In support of this, our current study results are generally concordant with the findings on the younger age groups studied in the WHI-CT. Another limitation is that we could not include venous thromboembolic events in our analyses owing to the absence of medical record confirmation of this outcome in the WHI-OS, which may have led to underestimates of total CVD risk with HT. Moreover, only a small percentage of women were using transdermal or low-dose estrogen, so the statistical power to assess these associations and to fully disentangle differences attributable to formulation versus dose was limited. Finally, because transdermal estrogen and low-dose estrogen are relatively newer HT formulations, their use may be a surrogate for unmeasured confounding health variables associated with lower CVD rates, such as type of prescribing healthcare provider, healthcare system, or other health habits or attributes of the participants.
We recognize that our observational cohort may underestimate HT-related CVD risk by missing events during the first few years of HT initiation for women who began HT before enrollment into the WHI study. Previous randomized studies have reported an increased CVD risk with HT use that is divergent from observational results. With regard to stroke, randomized trials and observational studies have been concordant, showing an overall increased risk with HT. Our study thus offers further insights into stroke risk based on findings according to estrogen formulation and route of delivery.