Comparing the Observational and Clinical Trial Arms of the Women’s Health Initiative
Special Report
By Leon Speroff, MD, Editor, Professor of Obstetrics and Gynecology, Oregon Health and Science University, Portland, is Editor for OB/GYN Clinical Alert.
The epidemiologic investigators of the Women’s Health Initiative (WHI) report a comparison of cardiovascular events in the observational arm with 53,054 women and the clinical trial arm with 16,608 women.1 The observational arm data were based on the 17,503 women using estrogen-progestin combination treatment. Overall, the incidence of coronary heart disease events and stroke was 50% less comparing the treated group to the placebo group in the observational arm, and venous thromboembolism (VTE) events were equivalent. Adjustment for confounding factors such as behavioral, dietary, physical activity, and cardiovascular risk factors left relative risks for coronary heart disease, stroke, and VTE still 30% to 38% lower than in the clinical trial (even though hormone users in the observational arm had less obesity). Following age adjustments, the relative risks were 39% to 48% lower. The authors believe their most important analysis was an adjustment for time from initiation of current hormone therapy. This analysis indicated that the hazard ratios comparing the 2 arms of the WHI were now in "fairly close" agreement. Indeed, they conclude that the "apparent discrepancy between the clinical trial and observational studies, such as the WHI observational study, may be substantially explained by classical confounding and differences in the distributions of time from estrogen-plus-progestin initiation."
This report represents a massive statistical exercise that prompts the accompanying editorial to state: "there is no overall benefit of hormone therapy in preventing cardiovascular disease," and "the conclusion that hormone therapy prevents cardiovascular disease, which was based on observational research, was wrong."2 This report from the WHI concluded that the hazard ratios "tended to decrease with increasing time from initiation of estrogen-plus-progestin use, and observational study hazard ratio estimates are heavily weighted by longer-term use while clinical trial hazard ratio estimates reflect shorter-term use."
But wait a minute. Let’s look at their numbers regarding time in years from initiation of estrogen-progestin therapy:
The authors calculated that this time adjustment for treatment now eliminated the apparent reduction in coronary events and stroke, bringing the conclusion that this comparison of the 2 cohorts yielded closely similar results. But look at the confidence intervals. The only statistically significant hazard ratios were an increase in coronary events with less than 2 years of treatment in the clinical trial arm, an increase in stroke with 2 to 5 years of treatment in the clinical trial arm but a 52% reduction in stroke with this same time period of treatment in the observational arm. Look at the case numbers. Can we accept the authors’ conclusions with case numbers of 5 to 27 in the observational arm? Despite this statistical manipulation, the authors could not demonstrate a statistically significant increase in cardiovascular events in the observational arm, except for venous thromboembolism.
The WHI report concludes that their analyses reinforce early elevations in cardiovascular disease risk among hormone users. In the adjudicated results from the cancelled estrogen-progestin arm of the WHI clinical trial, an increase in risk was present only in the first year of use, but in the subgroup analyses, only the women who were 20 or more years distant from menopause had a statistically significant increased risk of coronary heart disease (1.71; CI = 1.20-2.50).3 Subtracting this group from the rest of the participants, coronary heart disease now was observed in an identical prevalence comparing the treated and placebo groups.
It seems to me that the WHI has gone to extremes trying one adjustment after another to explain or even to eliminate the different results in the clinical trial and the observational study. Another, equally valid interpretation of the current report would be that that there is an increasing benefit in terms of protection against cardiovascular disease with increasing duration of exposure to hormone therapy. A case-control study in the United Kingdom found a significant reduction in the risk of myocardial infarction only with the use of hormone therapy for more than 5 years.4 Indeed, a trend for an emerging protection against coronary heart disease was observed in both arms of the WHI clinical trial with increasing duration of treatment.3,5
Critics have often cited the "healthy user effect" to explain the beneficial effects of hormone therapy in observational studies of cardiovascular disease. But this analysis of the WHI observational study adjusted for every conceivable measure of better health among the users and still the hormone users exhibited 30% to 38% lower risks compared to the clinical trial. In my view, this WHI report is an effective rebuttal of the healthy user criticism.
The authors comment that hazard ratios in the 2 cohorts did not differ substantially comparing women younger than 60 years of age at baseline and those older than age 60, or comparing women less than 10 years from menopause with women 10 or more years distant from menopause. But we know that the number of women in the younger age group and those close to the age of menopause was very small in the clinical trial. Is there sufficient statistical power to make this statement? In fact, in the adjudicated results from the clinical trial, the WHI reported that the test for trend was significant, indicating evidence of protection against coronary events in the younger women on hormone therapy.3
Another concern is the application of "adjudication." We know that when the coronary disease diagnoses were adjudicated by a central committee (with a change in 10% of the diagnoses of myocardial infarction), the increased risk for coronary events lost its statistical significance.3 The current WHI report indicates that only a fraction of the outcomes in the observational study were adjudicated. How big was that fraction, and unless a similar process and rate of adjudication are applied to the 2 studies, is this comparing apples and oranges?
There is one more important observation. In this report from the WHI, the hazard ratios for coronary heart disease and stroke in the clinical trial estrogen-progestin arm do not achieve statistical significance. When the magnitude of an effect is influenced from one report to another not only by the process of adjudication, but by various statistical adjustments, does not that mean that the strength of the finding is relatively weak, that the data are not robust?
I hope that in the coming months, we will see reviews from epidemiologists regarding the statistical exercises in this report. Until then, I am not convinced that the observational arm data indicating a reduction in cardiovascular events are not more in harmony with the general population of women for whom we prescribe hormone therapy (in contrast to the clinical trial arm results). The WHI overall results have been consistent with 20 years of research, except for the cardiovascular data. When corrected for age and state of health, there is reason to believe that the WHI cardiovascular data do agree with 20 years of research. Isn’t that more logical? A team of epidemiologists from the United Kingdom has argued that only cardiovascular disease is impacted by socioeconomic circumstances, and the observational studies on this subject have not controlled for this factor.6 But this current report from the WHI adjusted for every factor, including education, family income, and social functioning-and there still was a reduction in cardiovascular risk with hormone therapy in the observational arm, and evidence that the benefit increased with increasing duration of exposure.
References
- Prentice RL, et al. Combined postmenopausal hormone therapy and cardiovascular disease: toward resolving the discrepancy between observational studies and the Women’s Health Initiative clinical trial. Am J Epidemiol. 2005;162:404-414.
- Petitti DB, Freedman DA. Invited commentary: how far can epidemiologists get with statistical adjustment? Am J Epidemiol. 2005;162:415-418.
- Manson JE, et al. Estrogen plus progestin and the risk of coronary heart disease. N Engl J Med. 2003;349:523-534.
- Chilvers CE, et al. Post menopausal hormone replacement therapy and risk of acute myocardial infarction—a case control study of women in the East Midlands, UK. Eur Heart J. 2003;24:2197-2205.
- The Women’s Health Initiative Steering Committee. Effects of conjugated equine estrogen in postmenopausal women with hysterectomy. The Women’s Health Initiative Randomized Controlled Trial. JAMA. 2004;291:1701-1712.
- Lawlor DA, et al. Commentary: the hormone replacement-coronary heart disease conundrum: is this the death of observational epidemiology? Int J Epidemiol. 2004;33:464-467.
You have reached your article limit for the month. Subscribe now to access this article plus other member-only content.
- Award-winning Medical Content
- Latest Advances & Development in Medicine
- Unbiased Content