The negative association between E+P hazard ratio and time from menopause to first use of HT provides a possible explanation for a comparatively lower hazard ratio among women without prior HT in the clinical trial, since these women had much longer gap times.
Breast cancer hazard ratios in the observational study were in agreement with those from the clinical trial after controlling for both years from menopause to hormone therapy initiation and years since hormone therapy initiation (i.e., duration of E+P use for adherent women). For women who initiated E+P within 5 years of menopause—the group most likely making hormone therapy decisions in the future—the two data sources combined () to give hazard ratios of 1.85 (95 percent CI: 1.03, 3.34) for 2–5 years of use and 2.75 (95 percent CI: 1.73, 4.39) for more than 5 years of use. These analyses project an increase from 28 cases of invasive breast cancer per 10,000 person-years among nonusers of E+P, to 46 cases (attributable risk, 39 percent) over the first 5 years of use, to 61 cases (attributable risk, 54 percent) over the first 10 years of E+P use among women with gap times of less than 5 years.
Several biologic events could mediate a differential effect of E+P on breast cancer risk depending on time from menopause to initiation of HT. Preclinical studies indicate that breast cancers, when exposed to a period of estrogen deprivation, make adaptive changes (18
) that decrease their susceptibility to proliferative stimulation by estrogen (20
). In addition, combined hormone therapy increases mammographic density (21
) and slows the change from a dense pattern to a more fatty pattern, thought to represent lobular involution with reduction in the number of breast epithelial and stromal cells (23
). Because lobular involution is associated with reduced breast cancer risk (24
), a longer time from menopause with resultant lobular involution could decrease the number of epithelial breast cells potentially influenced by estrogen and progestin. Biologic inferences about E+P effects on breast cancer are somewhat limited by potential influence of these hormones on mammographic interpretation and breast cancer detection (2
Concerning data analysis methods, time from enrollment is the natural, basic time variable in Cox regression analysis of clinical trial data, but other choices may be of interest for cohort data analyses, including study subject age. Here, we defined time from enrollment as the basic time variable for both the clinical trial and observational study while stratifying breast cancer rates on baseline age. Doing so implies that hormone therapy hazard ratios derive from comparisons between E+P users and nonusers who are the same length of time from WHI enrollment and are also close in age. As such, these hazard ratios can be expected to be very similar to those that would derive from corresponding analyses that define age as the basic time variable (both clinical trial and observational study) that also stratify on baseline age (so that women of a given age during follow-up are also a similar time from enrollment and covariate ascertainment, within strata). For example, under this alternative modeling strategy, the E+P hazard ratios corresponding to the combined clinical trial and observational study analyses on the right side of are, respectively, 0.99 (95 percent CI: 0.56, 1.73), 2.05 (95 percent CI: 1.44, 2.92), and 2.96 (95 percent CI: 2.37, 3.68) for women without prior HTand 1.36 (95 percent CI: 0.70, 2.66), 2.44 (95 percent CI: 1.47, 4.07), and 3.33 (95 percent CI: 1.92, 5.79) for women with prior HT, while the hazard ratio factor for a 5-year gap-time increase is 0.81 (95 percent CI: 0.71, 0.91), in close agreement with those shown in .
Furthermore, the rather complex definition of prior HT status in the observational study may benefit from some elaboration. With an average baseline age of 63 years, there were few HT initiators during observational study follow-up, so that E+P user and nonuser groups were necessarily defined according to E+P use at enrollment. Women in the E+P user group had often used the study regimen for some years prior to observational study enrollment. Any use of another HT regimen prior to this ongoing baseline episode caused a woman to be classified as having prior HT. In addition, a woman who used the study regimen only prior to enrollment, but had a usage gap of 1 year or longer in this prior HT history, was classified as having prior HT. For such women, the duration of the ongoing baseline episode was the time from enrollment to the first usage gap of 1 year or longer encountered, going back in time.
The strengths of this study include the randomized controlled design of the clinical trial, with findings independently tested in the well-characterized observational study cohort. The two cohorts were drawn from the same populations; both received personal interviews regarding their history of hormone therapy use and had serial assessment of mammography use, common breast cancer risk factor assessment procedures, and very similar breast cancer ascertainment procedures. Study limitations include potential reliability issues associated with the retrospective assessment of both prior hormone therapy use (26
) and age at menopause (28
), especially among women who were many years past menopause at WHI enrollment. In addition, relief of vasomotor symptoms or risk of osteoporosis were likely reasons for observational study women to be using E+P at enrollment, whereas clinical trial women agreed to be randomly assigned to E+P or placebo. However, the good agreement between hazard ratios from the two cohorts suggests little, if any, hazard ratio confounding (i.e., effect modification) by this factor.
Another limitation relates to the few clinical trial women without prior HT having short gap times. This limitation implies that corresponding breast cancer hazard ratios from clinical trial analyses may be sensitive to modeling assumptions. For example, analyses of the type shown in , but with a 10-year gap-time cutpoint, do not provide evidence of a hazard ratio dependence on gap time. The analyses provide an examination of hazard ratios among women without prior HT that is rather robust to modeling assumptions, but some cells involved a small number of breast cancer cases, and most cases derived from the observational study in some cells. It will be valuable for the hazard ratio associations examined here to be considered in other studies, especially those that include many recent E+P initiators without prior HT, and that can estimate gap time and duration of E+P use with precision.
In summary, the WHI clinical trial and observational study each support an adverse effect of daily 0.625-mg conjugated equine estrogen plus 2.5-mg medroxyprogesterone acetate on breast cancer. Women who initiate treatment soon after menopause and continue for many years appear to be at particularly high risk.