Three More Reports from the WHI

Special Report

One year after the cancelation of the estrogen-progestin arm of the WHI, 3 more reports have emerged from this landmark study. Leon Speroff, MD, and Sarah L. Berga, MD provide their analysis as a benefit to OB/GYN Clinical Alert subscribers.

The May 28, 2003 issue of the Journal of the American Medical Association contained 3 reports from the Women’s Health Initiative (WHI), 1 on the risk of stroke from the canceled estrogen-progestin arm of the WHI, and 2 from the Women’s Health Initiative Memory Study (WHIMS).

WHIMS is a randomized, double-blind, placebo-controlled clinical trial that used the participants in the WHI to assess the risk of all-cause dementia, mild cognitive impairment, and "global cognitive functioning." The WHIMS subjects were all 65 years or older. In the assessment of dementia and mild cognitive impairment, 2229 women received estrogen-progestin (0.625 mg conjugated estrogens and 2.5 mg medroxyprogesterone acetate), and 2303 received placebo treatment. In the assessment of global cognitive function, 2145 received hormonal therapy and 2236 placebo. The results were reported after an average of 4 years of follow-up (see Table 1).

The WHI concluded that estrogen-progestin therapy increased the risk for probable dementia in women 65 years and older and did not prevent mild cognitive impairment. The WHIMS report on global cognitive function measured the participants annually with an assessment called the Modified Mini-Mental Examination. The results were expressed in rates of change and are best summarized by this conclusion: Cognitive function was no different comparing hormone therapy and placebo, but some of the women on estrogen-progestin treatment experienced a detrimental effect.

The WHI report on stroke was an extension of the initial report in July 2002, based on central adjudication of the diagnoses by neurologists. Eighty percent of all strokes were ischemic strokes; the small number of hemorrhagic strokes limited the power to make a conclusion regarding this type of stroke. The risk of ischemic stroke was increased with estrogen-progestin therapy (see Table 2) (Shumaker SA, et al. JAMA. 2003;289:2651-2662; Rapp SR, et al. JAMA. 2003;289:2663-2672; Wassertheil-Smoller S, et al. JAMA. 2003;289:2673-2684).

Comment By Leon Speroff, MD

The WHIMS trial was impressive in its extensive and rigorous detailed assessment of the participants, using a large battery of measures, as well as an adjudication committee consisting of 2 neurologists and 1 geriatrician that reviewed all diagnoses of dementia and a random sample of the cases of mild cognitive impairment.

The Report on Dementia and Impaired Cognition

A very important demographic characteristic of the WHIMS report is the age of the subjects: 47% were 65-69; 35% were 70-74; and 18% were 75 or older. It seems to me that the critical analysis is revealed in the statistical assessment by age group (see Table 3).

The only statistically significant finding was increased dementia in elderly women (75 and older) who had been exposed to a relatively short term of estrogen-progestin therapy. After excluding the 265 women at higher risk for developing dementia according to the baseline assessment, the case numbers were down to 24 in the treated group and 10 in the placebo group (HR = 2.64; CI, 1.26-5.53). How many of this small number were 75 or older? Indeed, the report recognized that cases of dementia appeared in the first year of treatment, suggesting that some of the women had cognitive problems at baseline. The report then concluded that hormone therapy hastened the progression to dementia in these women, but in my view it is difficult to make such a conclusion with such a small number of cases.

The average length of follow-up was 4 years, but only 50% of the subjects remained on hormone therapy for 4 years. The report on dementia stated that controlling for adherence did not alter the findings. However, it was revealed that after adjusting for nonadherence, the number of dementia cases was reduced to 21 in the treated group and 6 in the placebo group. The report on cognitive function concluded that after adjustment for adherence, the more favorable performance in years 3-5 in the placebo group was now only marginal, and thus, not clinically different. These analyses are important because the differences between the treated and placebo groups are not large (in the dementia analysis, the differences were 2, 5, 5, and 8 cases in years 1 through 4, and in year 5 the difference was 5, but this time the excess was in the placebo group). When case numbers are small, a shift in a few cases can change the conclusion. For this reason, it seems to me that adjudication of final diagnoses should be completed before presentation of the results to the public. The central adjudication process resulted in 20% disagreement in the treated group and 24% in the placebo group. The report states that most disagreements resulted in a less serious classification, but the effect on the final conclusions was not shared with us.

The Report on Cognition

The women in this report shared the same demographic characteristics as the report on dementia. The number of subjects that provided the conclusion that there was a small increased risk of a clinically meaningful decline in the hormone-treated group amounted to 6.7% of the women receiving hormone therapy. The confidence intervals are illustrated for these women, and the CIs are very wide, indicating small numbers in these groups. The clinical meaning for all postmenopausal women derived from these few women who experienced a detrimental effect of hormone therapy is by no means certain. What was different about this group of women?

The Report on Stroke

The difference in ischemic stroke was limited to nonfatal cases. An analysis of risk factors could not identify any unique characteristics that would indicate a higher risk group of women. An analysis adjusted for adherence found a similar increase in risk for overall stroke; HR = 1.50 (1.08-2.08). As in the first report from the WHI,1 the results were re-calculated for multiple end points with the Bonferroni adjustment, producing a confidence interval that was no longer statistically significant (0.93-1.84). What does this mean? Frankly, it is difficult for me to translate this into clinical application. I think Paul McDonough expressed it best when he considered the WHI and said that the statistical results are somewhere between the reported nominal numbers and the reduced numbers after adjustment.2 If that is the case, is the conclusion clinically meaningful?

The WHI reported that the increased risk of stroke with hormone therapy was observed in women with and without vasomotor symptoms (13 cases in the treated group and 6 in the placebo group in women with vasomotor symptoms), implying that results in women younger than 60 were similar, even in women with vasomotor symptoms. However, even the WHI admits that the number of events in women with symptoms was small, and we know from a recent WHI publication,3 that only 250 women aged 50-54 were on hormonal treatment. The weakness of the statistical conclusions is emphasized when the analyses of stroke events according to age groups or years since menopause all failed to reach statistical significance. The power to acquire significance was achieved only by lumping all strokes together.

These WHI reports suggest that the central nervous system findings (specifically, that with dementia) are related to vascular changes (specifically, silent strokes causing brain damage). This is a reasonable speculation, but it remains conjecture, especially because the statistical conclusions are limited in their strength.

Answered Questions

1. Will hormone therapy prevent subsequent clinical events in women with existing coronary artery disease? With a great deal of confidence, we can answer no, based upon the series of secondary prevention trials published over the last few years.

2. Will hormone therapy prescribed to elderly women prevent the appearance and progress of dementia? The results of the WHI and a secondary prevention trial in women with Alzheimer’s disease4 indicate that the answer is no.

Unanswered Questions

1. Will older women who have used hormone therapy for long durations early in their postmenopausal years be protected against dementia? The WHIMS report recognizes, in the discussion, that this hypothesis cannot be tested in this clinical trial. Nevertheless, the conclusion states that hormone therapy should not be prescribed with the expectation that it will enhance cognitive performance in postmenopausal women. It would have been more accurate to limit that statement to "postmenopausal women 75 years and older."

2. Will hormone therapy of long duration initiated early in the postmenopausal years have a beneficial effect on cognition? The WHIMS also cannot answer this question. A prospective study of a homogenous population in Utah (thus minimizing, if not eliminating, the healthy user bias) concluded that a reduction in the risk of Alzheimer’s requires long-term treatment, initiated at least 10 years before symptoms of dementia appear.5

3. Will hormone therapy reduce the risk of coronary artery disease when administered in a truly primary prevention fashion (for relatively long durations in women of early postmenopausal age with no clinical evidence of atherosclerosis)? Indeed, experimental evidence in women and monkeys indicates that as vascular endothelial cells become involved with atherosclerosis, beneficial responses to estrogen diminish.6,7

4. Will the adjudication process produce a shift in cardiac disease case numbers, eliminating the statistical significance of the conclusions already reported by the WHI?

An Emerging Theme

The controversies and recent results from the WHI suggest that it takes healthy target tissue to respond to hormone therapy. If cardiovascular epithelium progressively loses its ability to respond to estrogen in the presence of increasing atherosclerosis, and if neurons lose their ability to respond to estrogen in the presence of the pathology of Alzheimer’s disease, then the reported results (lack of effects in secondary prevention) are not surprising, and a truly primary prevention trial may yield a benefit of hormone therapy. The WHI has not tested this hypothesis because it has not studied the appropriate population—the population that forms the great majority of patients prescribed hormone therapy.

The WHI Has Many Limitations

Unfortunately, the individuals involved with the WHI, as well as uncritical supporters, make the leap from conclusions regarding a select group of women to generalizations applied to all postmenopausal women. It is repeatedly stated that the combined WHI results indicate that the risks of estrogen plus progestin outweigh the benefits. In my view, this is true for the specific sets of women studied by the WHI. Important questions remain unanswered, and therefore, hormone therapy initiated in early postmenopausal women and treatment maintained for many years are not precluded by the WHI results.

I continue to emphasize a clinical approach that I believe is acceptable and comforting for patients. Assist each patient to identify her specific goals and objectives. Once identified, an informed discussion can appropriately review the various treatment options to achieve those goals and objectives. This is a process to be repeated at least once a year, incorporating each time the available new knowledge. Approached in this fashion, terms such as short term or long term become meaningless.

Dr. Speroff is Professor of Obstetrics and Gynecology at Oregon Health Sciences University in Portland and editor of Ob/Gyn Clinical Alert.

Comment by Sarah L. Berga, MD

The WHIMS study results were unexpected and difficult to reconcile with prior data. While it is typically the case that most study results raise more questions than they answer, this study would seem to be all questions and no answers. What troubles me the most about this study is that there were only 32 cases of Alzheimer’s dementia despite the participation of 4532 women older than 65. However, only 803 of the participants were older than 75. This may explain why there were so few cases of AD. As the Cache County study5 so amply illustrated, rates of dementia do not start to climb in women until after age 80 years. In this study, the risk of developing probable dementia was 12.2 times greater for women aged 75 to 80 years than for those aged 65 to 69 years. Also, as the authors point out, cumulative hazards ratios began to diverge within the first year. Given the small number of cases, one has to wonder if there was a failure of randomization. This suspicion is heightened by the observation that the rates for mild cognitive impairment did not differ between the 2 arms.

Further, if one looks at Table 5 on page 2658 of JAMA, 35 of the women diagnosed with probable dementia had not used hormone therapy in the past as contrasted with 19 women in the placebo arm, giving a statistically significant different hazards ratio of 1.98 (1.13-3.47) for not having used HRT. Interestingly, this hazard ratio is similar to that for the risk of probable dementia (of all types) of 2.05 (1.21-3.48) for HRT users in the WHIMS. Does this historical feature explain the difference in dementia rates between the 2 arms? It is difficult to say. There were many variables (shown in Table 5) that differed between the 2 arms. If nothing else, the study illustrates that randomization alone does not guarantee that all relevant attributes will be evenly distributed between the 2 arms. Apparently, there is no study design that can completely obviate bias. I think that we would all agree that the presence or absence of prior hormone therapy is relevant to interpreting the results, but since this variable was not explicitly used as a factor in the randomization process, it is perhaps not completely surprising that the 2 groups differed in their past use of hormones. However, this asymmetry in HRT use is troubling, as the findings of the Cache County study showed that for hormone use to be neuroprotective, it had to be started around the time of menopause and continued for about 10 years. Further, since there were 88 cases of dementia in the Cache County study vs 61 in WHIMS and since having more cases gives greater confidence in the results, one wonders if one ought not to place more stock in the Cache County findings.

If the study results are questionable, where does that leave us, the physicians who must interpret the results to our patients? I was and am reluctant to put much faith in the results, but the headlines on the front page of the newspaper in my town were large and clear: "Hormone Use Doubles Risk of Dementia." One clearly has a lot of explaining to do even if one doesn’t trust the results. It is very hard to take a dismissive stance without looking like an intellectual dinosaur. On the other hand, most patients don’t really want a thorough statistical explanation of why the results can’t be trusted. I found it ironic that Palmer and colleagues just published a study in which they tested various diagnostic tools used for the diagnosis of dementia in the preclinical phase and found that only 18% of 1435 persons who ultimately developed dementia were identified using a 3-step process similar to that used by WHIMS.8

The WHIMS results might be real. There might be reasonable biological explanations for why the risk of dementia was doubled in the HRT-treated vs placebo group. In terms of putting the results into perspective, it would help a lot to know what the risk looks like in the estrogen-only arm. However, I am trying to resist the temptation to proffer a biological explanation when I feel like the results may not be "real." I am uncomfortable with offering a biological rationale when the findings may not stand the test of time. It will be impossible to completely ignore these study results. But it is also difficult to base therapy on them. My primary concern is that the results will actually guide clinical decision-making, that estrogen will be found to be neuroprotective, and that many women will decide to stop using HRT in any form because there is so much confusion.

Dr. Berga is Professor and Director, Division of Reproductive Endocrinology and Infertility, University of Pittsburgh.


1. Rossouw J, et al. JAMA. 2002;288:321-333.

2. McDonough PG. Fertil Steril. 2002;78:951-956.

3. Hays J, et al. N Eng J Med. 2003;348:1839-1854.

4. Mulnard R, et al. JAMA. 2000;284:1007-1015.

5. Zandi P, et al. JAMA. 2002;288:2123-2129.

6. Herrington D, et al. Arterioscl Thromb Vasc Biol. 2001; 21:1955-1961.

7. Mikkola TS, Clarkson TB. Cardiovasc Res. 2002;53: 605-619.

8. Palmer R, et al BMJ. 2003;326:245-247.