Controversy Over WHI Results

CONTROVERSY OVER WHI RESULTS

There is some controversy in the medical community about whether or not the results of the WHI were ‘valid’ or ‘relevant’, particularly in the context of divergent findings from earlier observational work and smaller clinical trials. Like most trials, the WHI may have had some minor limitations, which have been examined in an attempt to ascertain whether such problems might have biased the results suffi-ciently to change the direction of the effect of HT on CHD and on the global risk/benefit assessment. One of these limitations was the relatively high dropout and crossover rates that the study had. If dropouts and cross-over are differential, they would have had potential to bias the results of the study, particu-larly if dropouts or crossover occurred differentially based on women’s health status, which is plausible. However, these issues should not have biased the ‘intention-to-treat’ analyses.

Another concern voiced in the literature was the age of the WHI participants in the hormone trials compared with the average age of menopause. In most observational studies, HT is started at, or close to, the time of menopause. In WHI, as in most other clinical trials, the therapy was often initiated more than a decade, on average, after menopause, with an average age of women in the E & P trial of 63.2 years and of 63.6 years in the E-alone trial (WHI, 1998, 2002, 2004). If the effects, both positive and negative, of HT vary depending on the age of the women (or the duration of time since menopause), as has been suggested, conducting the study in older women may have biased it away from seeing favourable effects and towards seeing increased risks associated with HT in high-risk women. Also, baseline absolute rates of disease are much lower in recently menopausal women than in older women, so the absolute number of excess events will also differ by age.

One issue that spurred much debate was the use of what the WHI investigators termed a ‘global index’ – a composite measure consisting of the earli-est occurrence of CHD, invasive breast cancer, stroke, pulmonary embolism, endometrial cancer, colorectal cancer, hip fracture and death due to other causes – which was used to quantify the overall risk associ-ated with the drug (WHI Writing Group, 2002). Issues raised regarding the global index include the criticism that it was not validated (Goldman, 2004); the selec-tion of events to be included in the global index; and the index’s apparent lack of weighting of treatment types (despite very different health, quality of life, and economic impacts of each type of event).

Whatever the concerns, however, the findings (which seemed to contradict long-standing dogma), caused the medical and lay communities to express dismay and disbelief. As Elias Zerhouni, director of NIH, said about the HT controversy when WHI results were first published in 2002 – ‘Often in science the reaction to a new finding is directly proportional to the strength of the dogma it overturns. People are still in denial of the theory of relativity, too’ (Rothenberg, 2005; Spake, 2002). The way in which some of the WHI data was presented in the media may also have increased the concern expressed by some. There was an emphasis on Relative Risk (RR) versus rate differ-ence (RD), which may have served to magnify the risks observed and led to misinterpretations of the results. For example, while the relative hazard of CHD was elevated by 29% for women taking E & P, in absolute terms this equates on only an additional 7 cases of CHD per 10 000 women per year (WHI, 2002). Regardless of these concerns, however, the fact remains that WHI is still the largest random-ized clinical trial of HT to date, with the best ascer-tainment of outcomes. As such, the results of this landmark study should not be dismissed. As noted by the European Menopause and Andropause Society (EMAS) and by S. Barton, ‘it is not an easy task to opt between results of observational and clinical trials. High-quality observational studies may extend evidence over a wider population and are likely to be dominant in the identification of harms’, but the ‘best RCT still trumps the best observational study’ (Barton, 2000; Neves-e-Castro et al., 2002).