The Blunder They Called a Refutation

How evidence-based medicine tested two patented molecules and proudly buried a physiology.

Every medical student in the last twenty years has been taught the same lesson. There was a time, they are told, when hormone replacement therapy was prescribed enthusiastically. It was seen as protective against heart disease, against bone loss, against cognitive decline, against the wear of ageing itself. Then, in 2002, the Women’s Health Initiative was published in JAMA, and everything changed. The trial found that women taking combined hormone therapy had more breast cancer, more strokes, more venous thromboembolism, and more cardiovascular events than women on placebo. The data safety monitoring board halted the study early. The medical world corrected course. Prescriptions collapsed within eighteen months. Generations of physicians learned to describe hormone replacement as a cautionary tale, a case study of what happens when biological plausibility is not disciplined by randomised evidence.

This is the exemplar. It is taught in pharmacology, in gynaecology, in evidence-based medicine courses. It is the answer given when a student asks how EBM justifies its authority. WHI is presented as the founding refutation of a specific way of reasoning: the biology-based inference that women losing hormones should have their hormones restored. Someone reasoned that women losing estrogen should be given estrogen; someone else reasoned that a progestogen was needed to protect the endometrium; the whole cascade of biological plausibility produced a treatment that killed people. Only trials, EBM says, can discipline this kind of thinking. WHI is the case that discipline was administered. Biology was checked against evidence and found wrong.

Any argument that tries to rehabilitate hormone replacement on the strength of physiological reasoning is now read, by anyone trained in the current paradigm, as exactly the kind of undisciplined biology-first thinking EBM was created to correct. That reflex is not baseless. Undisciplined biological reasoning exists and can produce harm. But WHI is not what corrected it. WHI corrected something else. And the exemplar has been telling the wrong story about what it corrected for two decades.

What the pill contained

The Women’s Health Initiative did not test hormone replacement in any physiological sense. It tested two specific pharmaceutical products.

The estrogen arm administered Premarin. Premarin is the brand name of a mixture of conjugated estrogens extracted from the urine of pregnant mares. The letters CEE, conjugated equine estrogens, describe what it is: an equine biological product, containing estrone sulfate, equilin sulfate, 17α-dihydroequilin, and roughly a dozen other equine metabolites, in proportions that reflect horse pregnancy rather than human physiology. It contains almost no estradiol, which is the estrogen the human ovary secretes and the human body has been using for its entire endocrine life. It is delivered orally, which routes it through the first pass of the liver, generating a hepatic protein cascade that no ovulating woman ever experiences. It was patented by Wyeth in 1942 because Wyeth had industrial access to mare urine. The molecule that pharmacies dispensed for sixty years was chosen for supply chain reasons.

The combined arm added Provera. Provera is the brand name of medroxyprogesterone acetate, MPA. MPA is a synthetic modification of progesterone with a methyl group added to the six-position and an acetate at the seventeen. It is a progestin, not progesterone. Progesterone is what the corpus luteum secretes. MPA is a patented derivative developed to be orally active and durable, properties that the naturally secreted molecule lacks and that patents can protect. It binds progesterone receptors, but it also binds glucocorticoid and androgen receptors with an affinity that progesterone does not. It has been shown to alter breast epithelial proliferation in ways progesterone does not, to modify vascular reactivity in ways progesterone does not, to affect glucose metabolism in ways progesterone does not.

These are the two molecules the Women’s Health Initiative tested. A horse-urine extract, delivered by a route that bypasses the ovary’s physiology entirely. A synthetic derivative of progesterone chosen for its patent life. Neither is what the pre-menopausal female body produces. Neither has ever been claimed to be.

The category confusion at the heart of the claim

The word replacement carries a specific promise. It says that what is missing is being restored. If a diabetic is missing insulin, replacement means insulin, not an animal-derived mixture that binds some of the same receptors. If a thyroidectomised patient is missing thyroxine, replacement means levothyroxine, the molecule the thyroid secretes. If Addison’s disease is destroying the adrenal cortex, replacement means hydrocortisone, the molecule the adrenal makes. In every hospital in the Western world, replacement means the missing molecule in the form the body recognises.

Hormone replacement therapy, as tested in WHI, was not replacement in this sense. It was the administration of two patented molecules that shared some but not all receptor affinities with the missing endogenous hormones, delivered by routes that produced pharmacokinetics no ovulating woman has ever generated. To call this protocol replacement and to test it under that name was already to name it wrong.

The disciplined form of the physiological hypothesis is not that any molecule labelled estrogen will do. The disciplined form is that a woman whose ovaries have ceased functioning is now producing quantities of estradiol and progesterone that her tissues, formed over sixty years of higher exposure, are not adapted to. If restoration is attempted, it should use the molecules the body recognises, delivered by routes that generate the pharmacokinetics the body used to produce, in populations positioned to benefit. That is what a physiological reasoning worth defending actually says. What WHI tested departed from that hypothesis at every step. The molecules were chosen for supply and patent life. The dosing was chosen for regulatory simplicity. The population was chosen for logistical feasibility, most of them a decade or more past menopause. Every step at which the industrial protocol departed from the physiological hypothesis was a step made by industry and regulation.

The refutation was therefore of a scarecrow. The scarecrow wore lab coats and had a citation index. It looked, from outside, like the biology-based thesis. It was made of two patented molecules and two decades of marketing. The trial that took it down was, on its own terms, well conducted. The problem is not the trial. The problem is what the trial has been claimed to prove. It cannot, by any rule of experimental logic, refute a thesis that used different molecules, different routes, different timings, and different populations. The trial’s own protocol makes this explicit. WHI never enrolled women in early menopause on transdermal estradiol and cyclic oral progesterone. That regimen was not on the study arm. The claim that hormone replacement was refuted rests on the assumption that Premarin plus Provera is the same category of intervention as any other estrogen plus any other progestogen. That assumption is not a finding of the trial. It is a premise imported into its interpretation.

What EBM’s own literature published

The most efficient way to see the category error is to look at what evidence-based medicine itself has published on the molecules WHI did not test.

The E3N cohort, a French prospective study of nearly one hundred thousand women, was analysed by Fournier and colleagues in a series of papers between 2005 and 2008 in the International Journal of Cancer and Breast Cancer Research and Treatment. E3N compared users of different hormone therapy regimens, distinguishing bioidentical progesterone from synthetic progestins. The results, when disaggregated by progestogen, are stark. Estrogen combined with synthetic progestins including MPA and norethisterone produced a substantial increase in breast cancer incidence, of a magnitude consistent with the WHI signal. Estrogen combined with bioidentical micronised progesterone produced no statistically significant increase over the follow-up period. The Fournier group concluded that the association between combined hormone therapy and breast cancer depended on the type of progestogen. This is EBM’s own literature, published in EBM’s own journals, showing that the molecule identity matters.

The ESTHER study, published by Scarabin and colleagues in The Lancet in 2003 and elaborated over the following years, examined venous thromboembolism risk by route of estrogen administration. Oral estrogens, including CEE, increased VTE risk in postmenopausal women. Transdermal estradiol, which bypasses the liver’s first pass, did not. The mechanism is not controversial. Oral estrogens, whether equine or synthetic, upregulate hepatic synthesis of coagulation factors because they arrive at the liver in supraphysiological concentrations that no ovarian secretion produces. Transdermal estradiol delivers to the systemic circulation the way the ovary does, and does not.

Neither of these findings caused the exemplar to be retracted. The teaching continued. Medical students in the late 2020s are still told that WHI proved hormone replacement is dangerous, without these qualifications. The trial’s brand-level results are cited as if they applied to the class. The two decades of subsequent literature that dissected which molecules and which routes carry which risks are, in the exemplar as it is taught, invisible.

The exemplar that formed a generation

Between 2002 and 2020, prescriptions for menopausal hormone therapy fell by more than half in most Western countries. In the United States, the drop was steepest, and it happened within eighteen months of the WHI publication. Women in early menopause who might have benefited from restored estradiol were denied it. Women already on therapy were told to stop. Physicians who tried to distinguish between the trial’s specific findings and the population-level advice were characterised as ignoring the evidence.

Some of what followed is quantifiable. Bone density accelerated its decline in the affected cohorts. Hip fracture incidence rose. Cardiovascular endpoints, in the age-stratified reanalyses of WHI that appeared later, showed that women who began therapy near menopause had, if anything, cardiovascular benefit; the harm signal was concentrated in women who began therapy in their seventies, a decade or more after menopause, on a molecule that never resembled what their ovaries had made. The exemplar taught a rule that the trial’s own subsequent analyses undermined.

Other consequences are harder to count. The suffering of a decade of vasomotor symptoms, sleep disruption, cognitive fog, mood dysregulation, and sexual pain, borne by a generation of women who were told by evidence-based medicine that the treatment for their condition had been tested and refuted. The refutation, as documented above, was of a treatment they were not asking for. What they were asking for, restoration of a physiology they knew they had lost, had never been tested against placebo at scale and would not be for another twenty years.

The exemplar taught two generations of physicians to be sceptical of restoring what is missing, on the strength of a trial that never tested restoration. It taught women that what their bodies were signalling for was refused by evidence, on the strength of a trial that tested a horse-urine extract and a patented progestin. The refutation was proud. It was taught with the certainty of a paradigm secure in its methods. What it refuted was not what its critics were proposing. What it buried was the physiology it never named.

The Women’s Health Initiative was well conducted. Its findings are real for what it tested. The scandal is not the trial. The scandal is what it was claimed to prove, and the decades of denied replacement that followed from that claim.