"18,000 to 91,000 women. Who died because of a bad study."

Before we get into the data, I want you to sit with something.

There is no memorial wall for these women. No registry, no names, no obituaries that say “she died because her doctor stopped her estrogen in 2002.” Their deaths were filed under “cardiovascular disease” or “hip fracture complications” or “urosepsis.” They were statistical ghost deaths — real, countable, catastrophically preventable, and invisible.

The 50,000 to 150,000 women who may have died prematurely in the decades after the Women’s Health Initiative announcement? They weren’t identified as casualties. They were just women in their 50s who had bad luck.

Let’s talk about what actually happened.

The number everyone cites — and why precision matters more than a big round number

You’ve probably seen “150,000 women died.” It circulates in menopause advocacy spaces, in books, at conferences. It is not a peer-reviewed number. But it’s also not invented.

Here’s what Philip Sarrel, MD — an OB-GYN at Yale who spent his career in menopause medicine — actually calculated and published in the American Journal of Public Health in 2013:

Between 18,601 and 91,610 women died prematurely in the decade following the WHI announcement. Best estimate midpoint: approximately 48,000. TIME Magazine’s headline was “Could Estrogen Have Saved 50,000 Lives?” — and that’s the honest peer-reviewed figure.

The 150,000 number comes from extending Sarrel’s formula to the full 23 years from 2002 to 2025 — and expanding the population from his specific cohort to all women denied HRT, not just his carefully bounded group. HRT prescription rates dropped from 22% of postmenopausal women in 1999 to under 5% by 2010 — and they stayed there. The math is defensible. The number has not been peer-reviewed as a single figure. If you want intellectual precision, the right statement is: “At least 50,000 deaths in the first decade alone, in one specific population, with 23 years of suppressed prescribing still ongoing as of 2025.”

That’s less tweetable. It’s also more honest.

Okay so — who was Sarrel studying, and how did he get there?

What Sarrel actually found: the methodology, the population, the death causes

Sarrel focused on a specific population that made his case airtight: women who had hysterectomies, aged 50–59, in the decade following 2002.

Here’s why this population matters so much. Women without a uterus can take estrogen alone — no progestin required. And estrogen-alone was the arm of the WHI that actually showed a survival advantage for women aged 50–59 who took it. The all-cause mortality hazard ratio was 0.73 in favor of estrogen. If you’re in this population, you’re looking at data that says: “Women who took estrogen were 27% less likely to die in this window than women who didn’t.”

After the 2002 WHI announcement, doctors stopped prescribing estrogen to these women too. They threw out the signal along with the noise.

Sarrel took four data inputs — US Census population figures, hysterectomy prevalence rates (33–40% of women aged 50–59), the WHI’s measured mortality difference (13 excess deaths per 10,000 women per year on placebo), and the year-by-year decline in estrogen prescriptions (28% drop in 2002, growing to 70% by 2010) — and ran the math forward across every year from 2002 to 2011.

The range — 18,601 to 91,610 — comes from running the formula with the low and high ends of the WHI confidence interval for that mortality signal, combined with different estimates of how many women had hysterectomies and how sharply prescribing dropped.

What drove those excess deaths?

The paper is explicit: the mortality benefit of estrogen in this age group was “almost entirely owing to a decrease in coronary heart disease,” with additional contributions from reduced cancer mortality and other causes. These were primarily cardiac deaths — heart attacks — that estrogen’s cardiovascular protection would have prevented.

This is not a side effect story or a quality-of-life story. This is: estrogen protects the heart of a 50-year-old woman, you take it away, she dies of a heart attack five years later. The mechanism is that direct.

Making the number human: what 18,601 to 91,610 deaths actually means

Even the low end of that range — 18,601 deaths — is hard to make concrete until you put it next to something.

18,601 deaths is roughly the entire population of Breckenridge, Colorado, dying. In one decade.

The midpoint — 48,000 — is like the entire city of Flagstaff, Arizona. Gone.

91,610 at the high end: that’s every person in Santa Fe, New Mexico. Vanished from the map.

Another way to see it: 50,000 deaths over 10 years is 5,000 preventable deaths every single year, in this one population alone, for a decade. That’s thirteen women per day dying of something directly traceable to a policy decision made in response to a misread study.

Thirteen women. Every day. For ten years.

The women who died of 9/11 numbered 2,977. We built a memorial to each one of them. We know their names. We have their photos.

These 50,000 women — whose deaths were just as preventable, whose deaths were in fact predicted — have nothing. Not because their lives mattered less. Because the cause of their deaths was invisible even to their own families.

The Finnish data: what happens when women who ARE on HRT stop suddenly

Here’s what makes the human cost of 2002 even more acute. The women who were most immediately harmed weren’t just the ones who never got started on HRT. They were the women who were already on it, thriving, and then had their prescriptions discontinued by doctors who panicked.

A Finnish research group followed 432,775 women who discontinued hormone therapy between 1994 and 2013, tracking cardiac and stroke deaths. What they found in the first year after stopping is striking.

Among women under 60 who had been on HRT for more than five years and then stopped:

Cardiac death risk was 2.08× the expected population rate (SMR 2.08; 95% CI: 1.44–2.90)
Stroke death risk was 3.22× the expected population rate (SMR 3.22; 95% CI: 2.29–4.40)

In that 432,775-woman cohort, 5,204 women died of cardiac causes and 3,434 died of stroke during the study period. The risk was concentrated in the first twelve months after stopping — after one year, the excess risk diminished substantially.

Think about the timing. The WHI announcement was July 2002. Doctors panicked and pulled prescriptions. Millions of women who had been on estrogen for years — many of them in exactly the high-risk window — went cold turkey in August and September of 2002.

Women who had been protected by estrogen for five or more years, suddenly unprotected. Women under 60. Women for whom the risk of stopping was highest.

There’s also a dose-response pattern: the longer you’d been on HRT before stopping, the higher the first-year mortality spike. This makes biological sense — your cardiovascular system had adapted to estrogen’s protective effects over years; abrupt withdrawal was like suddenly removing a structural support.

The researchers found this elevated risk essentially vanished after the first year. Which means the acute harm was concentrated right in the months after the 2002 announcement, when millions of prescriptions were discontinued almost simultaneously.

Urosepsis: the UTI-to-death pipeline vaginal estrogen prevents

Most people hear “UTI” and think: annoying, three days of antibiotics, done.

Here’s what actually happens when a UTI spirals in a postmenopausal woman.

First: Bacteria colonize the urethra and bladder. In a woman without estrogen, the urogenital lining is thin and fragile. The vaginal pH has shifted from healthy acidic to alkaline, which means the lactobacilli that normally block bacterial colonization are largely gone. The bacteria find easy purchase and easy ascent.

Then: The bacteria breach the urinary tract and enter the bloodstream. At that point, you don’t have a bladder infection anymore. You have systemic sepsis — a bodywide inflammatory cascade that attacks the heart, lungs, kidneys, and blood-clotting systems simultaneously.

Then: In elderly women, this progresses fast. The mean time from urosepsis diagnosis to death in hospitalized patients is about eight days. The NIH estimates urosepsis mortality at 30–40%, escalating to 60% in severe cases with organ failure. The average age of the women who died in one prospective study was 76.4 years.

UTI → sepsis → death. In eight days.

This is the pipeline that vaginal estrogen interrupts at step one.

A 2025 study using more than two million electronic health records from multiple hospital systems compared women over 55 with recurrent UTIs who used vaginal estrogen versus those who didn’t:

Outcome	No vaginal estrogen	With vaginal estrogen
Sepsis rate	19.4%	10.6%
Mortality	1.54%	0.42%

Vaginal estrogen cut the sepsis rate nearly in half and the mortality rate by almost three-quarters.

That’s not a quality-of-life improvement. That’s a death-prevention intervention.

In a group of 1,000 postmenopausal women with recurrent UTIs, the difference between having vaginal estrogen available and not is roughly twelve fewer sepsis cases and eleven fewer deaths. Per thousand women.

And this is a local, low-dose intervention. The doses of vaginal estrogen needed to restore urogenital tissue are so small they don’t meaningfully enter the bloodstream. This is not systemic HRT. It’s a topical treatment for tissues that have estrogen receptors and know exactly what to do with it when it arrives.

Rachel Rubin, a urologist and sexual medicine specialist, stood before the FDA in July 2025 and said five words:

“Your label tried to kill my mother.”

Her mother was in the ICU. She needed vaginal estrogen to protect her from an ascending infection via her catheter. The nurses refused to give it. The pharmacists refused. The doctors refused. Not because vaginal estrogen was wrong for her mother. Because the black box warning on the packaging — inherited from 2003 WHI systemic HRT data — made everyone in the chain afraid to touch it.

The FDA removed that black box warning in November 2025. The label that nearly killed Dr. Rubin’s mother is gone.

But for 22 years, it was there.

The JAMA Oncology finding: the breast cancer survivor vindication

The deepest fear in the post-WHI era — the one that made even the most sympathetic doctors hesitant — was this: what if we give estrogen to a woman who has had breast cancer, and it feeds the cancer?

It’s a reasonable fear. Estrogen does stimulate estrogen-receptor-positive breast tumors. The concern has biological plausibility.

So here’s what the data actually shows.

In 2024, a study published in JAMA Oncology followed 49,237 women aged 40–79 with breast cancer — 84% of whom had estrogen-receptor-positive tumors — across Scotland and Wales, with median follow-up of up to eight years. The primary question: do women with breast cancer who use vaginal estrogen after diagnosis die from their cancer at higher rates than women who use nothing?

The result:

HR: 0.77 (95% CI: 0.63–0.94), p = 0.01.

Women with breast cancer who used vaginal estrogen had a 23% lower hazard of dying from their breast cancer compared to women who used no hormone therapy. The confidence interval excludes 1.0. This is statistically significant. This is not a trend; it’s a finding.

The authors were careful — as they should be in observational research. They framed it as “no evidence of increased mortality risk” rather than “protective.” Confounders are possible. Women prescribed vaginal estrogen might be healthier in ways that aren’t fully captured by the model.

But here is the plain logic: if vaginal estrogen were fueling fatal cancer recurrence, the hazard ratio should have gone above 1.0. Instead it went to 0.77. The fear predicted a number over 1.0 and got 0.77 instead.

The fear was backwards.

This doesn’t mean every breast cancer survivor should automatically be prescribed systemic HRT — those are different questions with different evidence. But for vaginal estrogen specifically, in a direct population study of nearly 50,000 breast cancer patients with long follow-up, the survival data does not support the fear that has kept oncologists from offering this treatment for two decades.

Approximately 55% of the women in the study were on tamoxifen and 48% were on aromatase inhibitors — standard-of-care breast cancer treatments that actively suppress estrogen. Even in that already-hormonally-suppressed population, vaginal estrogen did not make outcomes worse. It went the other direction.

The deaths are done. What matters now is the next woman.

The 50,000+ who died between 2002 and 2012 — we can’t reach them. The women who developed fatal arrhythmias in 2003 because their estrogen was stopped in 2002 are gone. The women who fell and broke a hip in 2008 and died of pneumonia in the hospital because their bone density collapsed after their prescription was cancelled — they’re gone. The women who progressed from UTI to sepsis in their 70s, in hospitals where nurses refused to restart vaginal estrogen — they’re gone.

We can’t count them by name. We can only count them by formula, by population data, by the cold math of what the WHI did to prescribing rates and what suppressed prescribing rates do to mortality.

But the next woman is still here.

The next woman is in her late 40s, feeling the first edges of perimenopause, Googling whether it’s safe to take hormones. She will find twenty years of fear-based content alongside the new science. She will walk into her doctor’s office and maybe — if she’s lucky — see a clinician who trained after the menopause education gap, who has read the 2023 NAMS guidelines, who understands the timing hypothesis and can counsel her appropriately.

Or she will see someone who still remembers the 2002 headlines and says “lowest dose, shortest time” as if it’s science.

The black box warning on vaginal estrogen is gone. The NAMS guidelines now explicitly state that for most healthy women under 60 who are within ten years of menopause onset, the benefits of hormone therapy outweigh the risks. The FDA’s own advisory panel voted in July 2025 to update the labels across the entire category. The reckoning is happening.

What you can do right now:

Know the Sarrel number. Not 150,000 — 18,000 to 91,000 deaths in ten years in one carefully bounded population, with the midpoint around 50,000. Peer-reviewed. Published in a major public health journal. Rebutted by the WHI investigators and defended by Sarrel. The debate is real; the direction of the evidence is not.

Know the Finnish number. If you’re currently on HRT and your doctor suggests stopping because “you’ve been on it long enough” — understand that stopping abruptly, especially if you’ve been on it for more than five years, carries a first-year mortality spike for cardiac and stroke death. Ask the question. Don’t just comply.

Know the vaginal estrogen number. Sepsis rate cut nearly in half. Mortality cut to one-quarter. In women over 55 with recurrent UTIs. This is a survival intervention that 22 years of mislabeling kept from elderly women in hospital ICUs.

Know the JAMA Oncology number. For breast cancer survivors asking about vaginal estrogen: the best available data shows HR 0.77 for cancer-specific mortality. Not 1.2, not 1.4. 0.77. The fear is not supported by the data.

The deaths from 2002 to 2025 are done. They’re in the ledger. They can’t be undone.

The woman who needs this information today — that’s who all of this is for.

Annette Thompson is 57, the founder of adoption.com, and a menopause advocate writing about evidence-based women’s health.

Sources: Sarrel et al., Am J Public Health, 2013 (PMID 23865654) | Venetkoski et al., Menopause, 2018 (PMID 29112596) | Mikkola et al., JCEM, 2015 | McVicker et al., JAMA Oncology, 2024 (PMID 37917089) | Wells et al., J Urology, 2025 (AUA abstract) | NIH StatPearls: Urosepsis | Healio, 2017: “Discontinuation of postmenopausal hormone therapy elevates cardiac, stroke death risk”

Keep reading

Get the research, not the reassurance.