More WHI answers than meet the eye

Few studies have made waves like the Women's Health Initiative's 2002 findings on hormone-replacement therapy. The 12-year clinical trial of more than 161,000 postmenopausal women found more risks than benefits for the study's hormone users — a surprising result in light of previous studies. But those discrepancies in findings become better understood when data from the WHI clinical trial and health-tracking study are analyzed together, according to new research by Public Health Sciences Division scientists.

Long-held assumptions about the benefits and safety of postmenopausal hormone use were mainly based on observational studies. This is research that tracks the health of participants to find patterns, but doesn't involve assigning women to hormone therapy or no hormone therapy, as a part of the study. Past observational research on postmenopausal hormone therapy reported a 40 percent to 50 percent reduction in heart disease incidence in women using these medications. In contrast, WHI's clinical trial found an elevated incidence of heart disease over its 5 1/2-year intervention period and stopped the estrogen and progestin part of the study early.

"The WHI results on hormone therapy and cardiovascular disease — especially coronary heart disease — were quite a surprise to many people because the clinical trial results differed so substantially from the preceding observational studies," said Dr. Ross Prentice, principal investigator of WHI's Clinical Coordinating Center.

Prentice and fellow PHS researchers Mary Pettinger and Dr. Garnet Anderson, co-principal investigator, sought to better understand the reasons for the discrepancies between the study's clinical trial results and earlier observational studies. Their paper, written with other WHI investigators, was published in the July 20 online edition of the American Journal of Epidemiology.

The researchers analyzed data from the 16,608 women in the hormone trial, along with corresponding data from the 53,504 women in the WHI observational study, 33 percent of whom were estrogen and progestin users at the start of the trial.

The scientists found the risk of heart disease, blood clots in the veins and stroke among women who used estrogen-plus progestin compared to women who do not was 39 percent to 48 percent lower in the observational study than in the clinical trial. The adverse effects of hormone use on these outcomes tended to occur early and then decrease.

Complementary results

The WHI clinical trial assigned women to hormone pills or to pills without hormones to assess risk from the beginning of hormone treatment. In most observational studies of hormone use, women were studied after they had used hormones for some years. Unfavorable effects are less apparent within a window of longer use. When the researchers took this short-term versus long-term difference in time since the initiation of estrogen plus progestin treatment into account, the risks were similar for hormone users and non-users, except for some unexplained difference for stroke.

"When we adjust for the factors that distinguish women who take combined hormones from those who don't and take into account the time period they'd already taken hormones previously, the results are fairly consistent between the WHI clinical trial and observational study," Prentice said. The researchers considered well-known risk factors for cardiovascular disease, such as age, race/ethnicity, body mass index, smoking history and age at menopause.

Purpose of paper

Prentice says the incomplete data available in previous observational studies to adjust for differences between estrogen plus progestin users and non-users reinforces the value of more rigorous clinical trials. "Some of the risk factors we accounted for are not routinely available in observational studies," he said. "Clinical trials, by design, balance participants in the treatment versus placebo arms of the study, taking into account all of these factors, plus some we might not even recognize."

Because the early WHI study results sparked such a profound sea change in medical practice — once common long-term postmenopausal hormone use plummeted — the trial has endured much scrutiny. "This paper is not so much a response to criticism as an attempt to bring together the existing data on these important topics," Prentice said. "It might indirectly stifle some criticism when people become aware that observational data, if collected and analyzed in a thorough manner, yields similar results to the WHI clinical trial."

Professional groups with interest in study design and data analysis issues have shown great interest in WHI's findings. Earlier this month, Prentice spoke about these results and other statistical topics related to WHI at the Joint Statistical Meetings in Minneapolis, the largest annual gathering of statisticians in North America.

In the world of epidemiology, Prentice says analyses like this show the importance of both clinical trials and observational studies. "Even though we were surprised by what was supposed to be a very strong hypothesis from an observational study viewpoint, the previous studies missed some important early effects," he said. "I think this analysis reassures us that we can come closer in our conclusions if we carefully take into account these other issues, like duration of use, as is already the case in some observational studies."

Importance of combined analylsis

"Our recent analyses certainly don't imply that there's not a need for clinical trials when you have topics of such public health importance. Rather, these analyses reinforce the complementary role that clinical trials and observational studies play. Instead of leaving us in a state of bewilderment, our combined clinical trial and observational analyses help us get organized in the important public-health research arena."

The study is funded by the National Heart, Lung, and Blood Institute and private donations. WHI researchers are currently analyzing the dietary intervention and calcium/vitamin D parts of the study and plan to release those findings in February 2006.