In 2002 a large trial was halted early, and the message that reached the public was that hormones cause breast cancer. The trial had two arms testing two different treatments, and they did not give the same result. This page sets out what the study reported, in its own numbers.
Reviewed by Allison Mollet, PA-C, MSCP
Updated August 2026
The Women's Health Initiative was among the largest studies ever conducted in women's health. In July 2002 its estrogen-plus-progestin arm was stopped early, and the finding was reported worldwide within hours.
The message that reached most women, and a great many clinicians, was that hormone therapy causes breast cancer. Prescribing fell sharply and did not recover.
The trial was well designed for the question it asked, and its investigators reported their results fully and accurately. What happened next was that a specific finding, in a specific population, on a specific drug combination, became a general belief about hormones.
The WHI hormone trials were prevention trials. They asked whether hormone therapy taken by women already years past menopause would prevent chronic disease later in life.
The participants reflect that purpose. The trials enrolled 27,347 postmenopausal women aged 50 to 79, and the average participant was 63 years old at enrollment. Only about a third were in their fifties (Manson et al., JAMA, 2017).
More telling still, most had no menopausal symptoms at all. In the estrogen-plus-progestin trial, 61% reported no vasomotor symptoms at enrollment, and fewer than 13% reported moderate or severe ones.
This was not a trial of treating menopause. It was a trial of taking hormones long after menopause to prevent disease. That is a reasonable question, and a different one from whether a 51-year-old with hot flashes and broken sleep should be treated.
The treatment matters as much as the population. The combined arm received oral conjugated equine estrogen with continuous medroxyprogesterone acetate, a synthetic progestin.
That regimen was chosen in the early 1990s from what was then most prescribed. It is not what a woman starting hormone therapy today would typically be given, and neither the route nor the progesterone is the same. Every figure below describes that combination, in those women.
Women with a uterus received the estrogen-plus-progestin combination. Women who had had a hysterectomy received estrogen alone. Two treatments, two separate trials, reported together in the public conversation.
The combined arm was stopped in July 2002. The estrogen-alone arm continued for another 19 months and was stopped in February 2004, for an increased risk of stroke rather than for breast cancer.
Over more than twenty years of follow-up, the difference became unmistakable.
Estrogen alone was associated with significantly less breast cancer. Not just less death from it — fewer cases. Incidence was lower than placebo (hazard ratio 0.78; 95% CI, 0.65–0.93), and so was breast cancer mortality (hazard ratio 0.60; 95% CI, 0.37–0.97).
In the combined arm, incidence was significantly higher (hazard ratio 1.28; 95% CI, 1.13–1.45).
But breast cancer mortality in that arm was not significantly different from placebo (hazard ratio 1.35; 95% CI, 0.94–1.95). More cancer was diagnosed. Over twenty years, women were not significantly more likely to die of it (Chlebowski et al., JAMA, 2020).
The signal that defined public understanding of hormone therapy came from the arm containing the synthetic progestin. In the estrogen-alone arm, women got breast cancer less often, and died of it less often, than women on placebo.
One limit the investigators drew themselves, and we draw it too: this is not a reason to take estrogen to prevent breast cancer. The benefit was concentrated in women without benign breast disease or a family history, and it does not transfer to women at raised risk (Anderson et al., Lancet Oncology, 2012).
"Hormones cause cancer" is not what this trial found. In the arm without the synthetic progestin, women had significantly less breast cancer than women on placebo, and were less likely to die of it. That is the opposite of the headline. Both arms were published, and only one became common knowledge.
— Allison Mollet, PA-C, MSCP
The original numbers are more specific than the conclusion drawn from them.
The breast cancer finding was a 26% relative increase. In absolute terms that was 38 cases per 10,000 women per year against 30 in the placebo group: eight additional cases per 10,000 women each year. Both figures describe the same result. Only the second tells a woman what she is deciding about.
By the trial's own standard, it did not reach statistical significance. The hazard ratio was 1.26, and the paper reports two confidence intervals around it: a nominal one of 1.00 to 1.59, and an adjusted one of 0.83 to 1.92.
The difference between them is that the trial examined its own results repeatedly as data accumulated, and the adjusted interval accounts for that. It includes 1.0, meaning no effect. The paper's own description is that the increase "almost reached nominal statistical significance."
The finding that changed how a generation of women were treated did not meet the threshold the trial had set for itself.
All-cause mortality was unaffected. Deaths from any cause were 231 in the hormone group and 218 on placebo, a hazard ratio of 0.98. The paper states it directly: "All-cause mortality was not affected during the trial."
There were benefits, and they were also measured. Six fewer colorectal cancers and five fewer hip fractures per 10,000 women per year. The 24% reduction in total fractures was statistically significant on both nominal and adjusted analysis, which the breast cancer increase was not.
The trial's own conclusion stands. Taking everything together, the investigators found that overall risks exceeded benefits for this regimen in this population, and that it should not be used to prevent coronary heart disease.
That conclusion was sound for what was tested. It is not the same statement as "hormones cause cancer," and the distance between the two is where twenty years of women's care was lost.
Source: Writing Group for the WHI Investigators, JAMA, 2002. Average follow-up 5.2 years. All-cause mortality was not affected (hazard ratio 0.98).
The WHI participants have now been followed for 18 years, and the mortality results were published in JAMA in 2017.
Across both trials, hormone therapy was not associated with any increase in death: not from all causes, not from cardiovascular disease, and not from cancer.
All-cause mortality was 27.1% among women who received hormone therapy and 27.6% among those who received placebo (hazard ratio 0.99; 95% CI, 0.94–1.03).
That is the summary measure that matters most, because it captures the net effect of everything the therapy does, beneficial and harmful together. Over nearly two decades, that net effect was neutral.
The trial also shows why timing matters, within its own data.
Among women who began treatment in their fifties, all-cause mortality during the treatment years was lower than among those given placebo (hazard ratio 0.69; 95% CI, 0.51–0.94). Among women in their seventies it was not.
Comparing the youngest group with the oldest, the ratio of hazard ratios was 0.61 (95% CI, 0.43–0.87), a statistically significant trend by age.
Younger women did better on hormone therapy than older women did. That pattern is why current guidance is built around a window rather than a blanket position, and why the population this trial enrolled least of is the population the guidance now addresses.
The 2002 announcement did not distinguish the trial's two arms. A result from one combination, in women averaging 63, was reported as a finding about hormone therapy.
Prescribing fell by roughly 80% within a year. It never recovered.
In that same trial, the reduction in fractures was statistically significant. The increase in breast cancer was not.
The finding that changed how women were treated for twenty years was the one that did not reach significance. The finding that did reach significance was a benefit.
The investigators reported significant heterogeneity between the two arms and declined to combine them. The public account combined them anyway, and kept combining them for two decades.
The correction is underway now. The data did not change in that time. Only the reading of them did.
A 2013 analysis in the American Journal of Public Health put a number on the cost.
Among hysterectomized women aged 50 to 59 in the United States, the authors estimated between 18,601 and 91,610 excess deaths in the decade after 2002. Their best estimate was near 40,000 to 49,000.
These were women who would have taken estrogen alone. The arm that showed benefit.
Their account of how it happened is the one this page has made. The trial drug was widely called "hormone therapy" or "estrogen," so the findings were generalized to estrogen alone in women without a uterus. Only women with a uterus had been randomized (Sarrel et al., 2013).
Breast cancer figures: Chlebowski et al., JAMA, 2020 — more than 20 years median cumulative follow-up, mortality data available for over 98% of the 27,347 women randomized. All-cause mortality: Manson et al., JAMA, 2017 — 18-year follow-up.
The WHI's two arms differed in one component: the synthetic progestin. That raises an obvious question about what is and is not known.
Observational studies suggest the type of progestogen used alongside estrogen may influence breast cancer risk. Micronized progesterone, structurally identical to the progesterone the body makes, has generally been associated with little or no increase in risk.
Certain synthetic progestins have been linked to a higher risk, similar to what the WHI reported.
No randomized trial has compared the two head to head. This is an important clinical consideration rather than a settled fact.
It is enough to inform a choice. We use micronized progesterone for endometrial protection in women with a uterus. It is the hormone the body produces, the observational data point in its favor, and there is no reason to prefer a synthetic alternative in its place.
The trial did not show that hormones are dangerous. It showed that one regimen, in older women well past menopause, carried risks that outweighed its benefits for the purpose of preventing disease.
What follows from that is timing, route, and which progesterone. Where a woman is relative to menopause changes the balance. Non-oral estrogen carries lower clot risk than oral. Micronized progesterone is what the observational evidence favors.
Our benefits of hormone therapy page covers what hormone therapy treats and protects, and who it suits.
The women who most stand to benefit from hormone therapy are the ones this trial enrolled least of. A trial in 63-year-olds without symptoms became the reason 51-year-olds with symptoms went untreated for two decades. The correction has been slow because the original headline was simpler than the data.
— Allison Mollet, PA-C, MSCP
Let's begin
Whether hormone therapy is a good option depends on your history, your symptoms, and where you are relative to menopause. An evaluation answers that with comprehensive labs and an unhurried hour with a Menopause Society Certified Practitioner. To speak with us first, call 714-671-2005.
The benefits of hormone therapy
The forms of hormone therapy, and how treatment is personalized
Menopause: what changes, and what helps
Perimenopause: what's happening, and what helps
The two arms gave opposite answers, and twenty years of follow-up made that clearer rather than murkier.
Women on estrogen alone had significantly less breast cancer than women on placebo, and were significantly less likely to die of it.
In the arm containing the synthetic progestin, more breast cancer was diagnosed. Deaths from breast cancer in that arm were not significantly higher than placebo.
The answer depends entirely on which hormones. That is the distinction the 2002 coverage collapsed, and it is the one that determines what we prescribe.
No. It was well conducted and reported accurately. Its conclusion, that this regimen's overall risks exceeded its benefits in this population and that it should not be used to prevent heart disease, was sound for what was tested.
What went wrong was applying it to women in their early fifties being treated for symptoms, and treating two different regimens as one.
No. All-cause mortality was not affected during the trial itself (hazard ratio 0.98), and over 18 years of follow-up hormone therapy was not associated with increased death from any cause, from cardiovascular disease, or from cancer. Among women who started in their fifties, mortality during the treatment years was lower than among those on placebo.
The only difference between them was the synthetic progestin. Women with a uterus received it, because estrogen without a progestogen increases the risk of uterine cancer. Women who had had a hysterectomy did not need it and received estrogen alone.
The investigators reported statistically significant heterogeneity between the trials and did not pool their breast cancer mortality results.
Observational studies suggest the type of progestogen may matter. Micronized progesterone has generally been associated with little or no increase in breast cancer risk, and certain synthetic progestins with higher risk.
No randomized trial has compared them directly, so this is a clinical consideration rather than a settled fact. We use micronized progesterone for endometrial protection in women with a uterus.
"Bioidentical" means the hormone is structurally identical to the one your body produces, and such hormones exist both as FDA-approved products and as compounded preparations.
The evidence supports three specifics rather than a category. Route matters, since clot risk is higher with hormones taken by mouth. So does the difference the WHI found between its two arms, and the observational signal favoring micronized progesterone.
A nominal interval treats the analysis as a single look at the data. An adjusted interval accounts for the trial having examined its results repeatedly as they accumulated, which makes a chance finding more likely and widens the interval.
The 2002 paper reports both for breast cancer. The adjusted interval includes no effect.
