The HRT-Prevents-Heart-Disease Doctrine — a 16,608-Woman Trial Halted for Causing the Heart Attacks

Summary

On 8 July 2002, the writing group of the Women's Health Initiative announced that it had stopped the estrogen-plus-progestin arm of its trial roughly three years early, after a mean of 5.2 years, because the combined hormone regimen taken by 16,608 healthy postmenopausal women had increased coronary heart disease rather than preventing it — a hazard ratio of 1.29, alongside elevated stroke (1.41), breast cancer (1.26), and blood clots (roughly doubled). The promise and the reality had been inverted: for nearly two decades, conjugated equine estrogen had been prescribed to protect women's hearts on the strength of observational data, and the first large randomized test of that promise in healthy women found the hearts were marginally worse off. The gap between the doctrine and the data would be counted in millions of prescriptions written for a benefit that did not exist.

The cardioprotection belief was not a fringe idea. It was mainstream preventive cardiology, taught in medical schools and embedded in clinical guidance, resting on some 40 to 50 observational studies — most prominently the Nurses' Health Study, which reported current HRT users with coronary heart disease risk reduced by roughly 40 percent (relative risk near 0.61). Conjugated equine estrogen, sold by Wyeth as Premarin since 1942 and as the estrogen-progestin combination Prempro from 1995, exceeded two billion dollars in sales in 2001. The molecules were never recalled; they remain licensed today for menopausal symptoms. What collapsed in July 2002 was the theory attached to them — the claim that they prevented heart disease — displaced not by a scandal or a fraud but by the single most decisive instrument in clinical medicine: a large, randomized, placebo-controlled trial pointed directly at the question the observational data had only ever circled.

The reversal had been foreshadowed. In 1998, the HERS trial — 2,763 women who already had coronary disease, randomized to estrogen-plus-progestin or placebo — had found no overall benefit for secondary prevention, and a 50 percent excess of cardiac events in the first year. The field absorbed HERS as a special case (sick women, late timing) and held to the primary-prevention promise. WHI tested that promise in healthy women, and it failed too. The market answered within months: combined-therapy prescriptions fell by roughly two-thirds within a year, and in 2003 the US Food and Drug Administration imposed a class-wide boxed warning.

This dossier files "Overturned" entry TH-006 as the archetype of a different failure than fraud. No one falsified data. A generation of physicians inferred causation from correlation in self-selected, healthier-than-average hormone users, and only a randomized trial could break the spell — long after the prescription pads had done their work.

Timeline

1942

Premarin licensed

Wyeth-Ayerst's conjugated equine estrogen (extracted from pregnant mares' urine) is approved for menopausal symptoms — decades before any cardiovascular claim.

1985

Observational cardioprotection takes hold

The Nurses' Health Study and a wave of cohort studies report large reductions (≈30–50%) in coronary heart disease among HRT users, hardening the prevention belief.

1991

NIH launches the WHI

Under director Bernadine Healy, the Women's Health Initiative is funded to test, by randomized trial, what observation had only suggested — including HRT and heart disease.

1995

Prempro arrives

Wyeth markets the fixed estrogen-progestin combination (CEE 0.625 mg + medroxyprogesterone 2.5 mg); progestin is added to protect the uterus, not the heart.

Aug 1998

HERS warns

The Heart and Estrogen/progestin Replacement Study (JAMA 1998;280:605–613; Hulley et al., 2,763 women with prior CHD) finds no secondary-prevention benefit and early excess events.

2000–2001

Early WHI signal

WHI's own data and safety monitoring board observes an early increase in cardiovascular events in the hormone arm; investigators send caution letters to participants.

31 May 2002

DSMB recommends stopping

After ~5.2 years, the monitoring board concludes the global-index harms exceed benefits and recommends halting the estrogen-plus-progestin arm.

8 Jul 2002

Public announcement

The WHI investigators announce the early termination; combined HRT is shown to raise CHD, stroke, clots, and breast cancer in healthy women.

17 Jul 2002

Principal results published

JAMA publishes the trial: CHD HR 1.29, stroke 1.41, breast cancer 1.26, VTE ~2.1, against fewer fractures and colorectal cancers.

2002–2003

Prescriptions collapse

Combined-therapy prescriptions fall by roughly two-thirds within a year; HRT use halves; Premarin/Prempro sales slide from $1.3B (2002) toward $880M (2004).

2003

FDA boxed warning

The FDA mandates a class-wide black-box warning on menopausal estrogen and estrogen-progestin products, citing the WHI cardiovascular and cancer findings.

2003–2007

The breast-cancer dividend

US invasive breast-cancer incidence drops sharply (notably in 2003), an effect researchers attribute substantially to the abrupt fall in HRT use.

Correlation Mistaken for a Cure

The cardioprotection doctrine was constructed honestly, which is why it lasted. By the early 1990s, dozens of observational studies — cohorts following hundreds of thousands of women — had converged on the same answer: women who took HRT had far less heart disease than women who did not. The Nurses' Health Study, the field's flagship, put current users at roughly 40 percent lower coronary risk. The mechanism was biologically plausible: estrogen improves lipid profiles and was assumed to keep arteries supple. From convergence, plausibility, and mechanism, the profession drew causation. What it could not see from inside the data was that women who chose and stayed on hormone therapy were systematically different — better educated, slimmer, more health-conscious, more likely to see a doctor, the so-called "healthy-user" and "compliance" effects. The therapy was a marker of women who would have had healthier hearts anyway. Observation cannot separate the drug from the kind of person who takes the drug; only randomization can. The structural lesson is that consistency across forty studies is not evidence of causation if all forty share the same confounder.

The Trial That Was Designed to Settle It

The Women's Health Initiative existed precisely because someone had recognized the gap. Bernadine Healy, the first woman to direct the NIH, championed a randomized test rather than another cohort, and in 1991 the WHI was funded to enroll tens of thousands of postmenopausal women and assign them, by chance, to hormones or placebo. Randomization is what neutralizes the healthy-user effect: it distributes the unmeasured differences evenly across both arms, so that any divergence in outcome can be attributed to the pill and not to the person. When the divergence came, it ran the wrong way. The 1998 HERS result was the first crack — no benefit in women with established disease, and a worrying first-year spike — but it was rationalized as a secondary-prevention artifact. WHI removed every excuse: healthy women, primary prevention, large numbers, placebo control, independent monitoring. When the data and safety monitoring board looked at the accumulating events in May 2002, the coronary curve was bending up, not down. There was no honest way to continue.

The Reckoning the Market Ran Faster Than the Journals

The reversal of belief was, for once, almost immediate — because the instrument was unimpeachable. Within nine months hormone-therapy prescriptions fell by roughly a third overall, and combined estrogen-progestin scripts by close to two-thirds; Wyeth's franchise, worth over two billion dollars in 2001, lost a substantial fraction of its volume inside two years. In 2003 the FDA closed the loop with a boxed warning across the drug class. And then the epidemiology produced a grim confirmation of the trial: US invasive breast-cancer incidence dropped measurably in 2003, a decline researchers tied to the sudden withdrawal of millions of women from HRT — evidence that the harm side of the ledger had been real all along. The aftermath was not simple vindication of the trialists, however. A later "timing hypothesis" reanalysis argued that younger women, treated near menopause, might see neutral or even favorable cardiac effects, and that WHI's older average enrollee (63) had overstated the harm for the women most likely to start therapy. The doctrine that HRT prevents heart disease in everyone did not survive; a narrower, hedged, age-stratified picture replaced it. The case is now the textbook byword for the limits of observational medicine.

Contributing Factors

01

Causation inferred from confounded observation

Forty-plus cohort studies agreed that HRT users had less heart disease, but users were self-selected healthier women. The "healthy-user" and "compliance" biases manufactured a protective signal that randomization later erased. Agreement among observational studies measures shared confounding as readily as shared truth.

02

Biological plausibility used as proof

Estrogen's favorable effect on cholesterol gave the doctrine a mechanism, and a plausible mechanism made the correlation feel causal. But plausibility predicts nothing about the net clinical balance; estrogen's pro-thrombotic and inflammatory effects, invisible to the lipid story, drove the early excess of events.

03

Early warnings absorbed as exceptions

The 1998 HERS trial — null benefit, first-year harm — was the doctrine's first randomized contradiction, and the field rationalized it away as a quirk of secondary prevention rather than a signal about the primary-prevention claim. A disconfirming result explained away is a reversal deferred, not avoided.

04

Prevention prescribed to the healthy at scale

Because the claimed benefit was prophylactic, the therapy was given to millions of asymptomatic women for years, multiplying any iatrogenic risk across an enormous, low-baseline-risk population. When a preventive is wrong, the harm is dispensed precisely to people who were not sick to begin with.

05

Commercial entrenchment of an unproven indication

A two-billion-dollar franchise, decades of marketing, and guideline endorsement gave the doctrine institutional mass and inertia. The molecules were legitimate for symptom relief, but the lucrative cardioprotection rationale extended their use far beyond what evidence supported, and the money made the belief harder to revise.

Aftermath

The material consequence was swift: HRT use roughly halved, combined-therapy prescriptions fell by about two-thirds, and Wyeth's Premarin-Prempro franchise contracted from $1.3 billion in 2002 toward $880 million by 2004, with the FDA's 2003 boxed warning sealing the retreat. The durable ripple was epistemic. WHI became the standard case study in why evidence-based medicine privileges the randomized controlled trial over observational cohorts, however large or consistent — the definitive demonstration that the "healthy-user" effect can fabricate an entire preventive doctrine. The measurable 2003 drop in US breast-cancer incidence stands as the population-scale receipt for the trial's harm findings. What remains is more nuanced than a clean refutation: the later timing hypothesis carved out a defensible role for hormone therapy in younger, recently menopausal women for symptom relief, and the boxed warning itself was eventually re-litigated and removed for menopausal products in late 2025. But the specific claim that revoked itself here — that HRT protects the postmenopausal heart — did not return. "Overturned" files this as TH-006 because it is the family's purest specimen of a true belief that was never true: not fraud, not malice, but a profession reading causation into a correlation and prescribing it to millions until a randomized trial, pointed at last at the right question, returned the opposite answer.

Lessons

Treat consistency across observational studies as a hypothesis, not a verdict — forty cohorts sharing the same healthy-user confounder will agree forty times over and still be wrong; reserve causal claims for randomized evidence.
Do not let a plausible mechanism substitute for a measured outcome: estrogen's lipid benefit was real and the net cardiac effect was harmful, so demand the clinical endpoint, not the surrogate that flatters it.
When an early randomized result contradicts a cherished doctrine, investigate the doctrine, not the exception — HERS warned in 1998 and was explained away, costing four more years of unwarranted prescribing.
Apply the highest evidentiary bar to preventives given to the healthy at scale, because a wrong prophylactic dispenses its harm to people who were never sick and multiplies it across the entire treated population.
Separate a drug's legitimate indication from its profitable one — the molecules were defensible for symptoms and indefensible for cardioprotection, and commercial momentum behind the second kept the error alive long past its evidence.