EBM: Evidence-Biased Medicine

An Essay on the Machinery That Decides What Counts as Knowing

The basis was biased, so the bias became the basis, and the basis became the base on which evidence-based medicine was based.

The 1992 Inversion

In November 1992, the Journal of the American Medical Association published a paper titled “Evidence-Based Medicine: A New Approach to Teaching the Practice of Medicine.”¹ The authors, the Evidence-Based Medicine Working Group at McMaster University, led by Gordon Guyatt, announced a paradigm shift. The first paragraph named what was being replaced: “intuition, unsystematic clinical experience, and pathophysiologic rationale.”¹ The replacement was a hierarchy in which the randomised controlled trial sat at the top and clinical observation sat near the bottom.

The paper was not modest. It described its proposal in Kuhnian terms and predicted that the old approach — the physician’s accumulated judgement, the recognition of patterns across thousands of patients, the reasoning from mechanism and first principles — would be superseded.¹ Within a decade, the framework had been adopted across major medical journals, accreditation bodies, and clinical guideline organisations. In a 2007 BMJ poll of more than 11,000 readers asked to name the most important medical milestones since 1840, the sanitary revolution placed first, antibiotics second, anaesthesia third; evidence-based medicine appeared on the shortlist of fifteen.²

What was elevated to the top of the hierarchy was the one form of evidence pharmaceutical companies could afford to manufacture at scale. What was demoted to the bottom was everything they could not control. This was not the discovery of how medicine should be practised. It was the redefinition of what counted as knowing. The framework called itself evidence-based. What it actually was, was evidence-biased — a hierarchy in which what counted as evidence was determined, first, by who could afford to produce it.

The essay examines what that redefinition did, who it served, and the cost in lives.

Support This Work

This work remains free because paid subscribers make it possible. If you find value here, consider joining them.

Paid subscribers get access to all books — including The DMSO Book, The Kitchen Remedies Guide, Chlorine Dioxide, The PSA Trap, Breast Cancer, and more — with 1-2 new books added each month. Plus the Deep Dive Audio Library: 180+ in-depth audio book summaries and discussions.

Paid subscribers also get the Questions for Your Doctor series, short printable guides built around the questions doctors are not trained to answer, one decision at a time. The companion Before You Consent series goes deeper, walking through a single diagnosis or prescription from the moment it lands to the days after, with every figure sourced. And the Package Insert series reads the labels the doctor never opens and the pharmacist never mentions, what the manufacturer admits, in their own words, about what they sell.

I do this work independently, outside the institutions these books so often describe. No foundation grants, no academic approval, no editorial gatekeepers deciding what’s acceptable to publish. Your subscription makes that independence possible.

---

Explain It to a Six-Year-Old

For a long time, children have learned about the world in many ways. Some things they see with their own eyes. Some things they hear from grandparents who have lived a long time. Some things they figure out by thinking carefully. Some things they know because they have tried them and watched what happened. All of these are ways of knowing.

One day, the school makes a new rule. From now on, the only things that count as knowing are things that have been seen in a special room, by a man with a clipboard, who writes down what he saw. Hearing from grandparents does not count anymore. Trying things and watching what happens does not count. Thinking carefully does not count. The teacher tells the children, “Those were just stories. Real knowing happens in the special room.”

The special room is very expensive. Only one company can afford to rent it. The company pays the man with the clipboard. The company decides what gets looked at in the room and what does not. Apples never go in the room. Sunlight never goes in the room. Grandmothers’ soup never goes in the room. So the school says, “We do not know if apples are good. We do not know if sunlight is good. We do not know if grandmothers’ soup is good. Nobody has seen them in the room.”

The company sells biscuits. The biscuits go in the room every day. The man with the clipboard writes down that the biscuits are good. The teacher tells the children, “We know the biscuits are good, because we saw them in the room.”

Some children eat the biscuits and get tummy aches. They tell the teacher. The teacher says, “Tummy aches have not been seen in the room. We do not know that the biscuits cause tummy aches.” When a doctor visits and says she has seen many children with tummy aches after eating biscuits, the teacher says the doctor is only telling stories. Stories do not count.

Years pass. The children eat more biscuits and fewer apples. Many of them are sick. The company is very rich. The teacher still says the only real knowing is the knowing that happens in the special room.

That is what happened to medicine in 1992.

---

What Sits at the Top of the Hierarchy

The EBM hierarchy of evidence places systematic reviews of randomised controlled trials at the apex. Beneath them sit individual RCTs, then cohort studies, then case-control studies, then case series, then expert opinion and clinical experience at the bottom.³

The framework is presented as neutral. It is not.

A randomised controlled trial of a pharmaceutical product costs between twenty and three hundred million dollars to conduct.⁴ The trial requires regulatory approval, site recruitment, statistical infrastructure, monitoring, data management, and publication support. The entities capable of funding such trials are, in practice, three: pharmaceutical companies, the National Institutes of Health, and a small number of large foundations whose priorities track institutional medicine. The pharmaceutical industry funds the majority of clinical research in the United States and a higher proportion of late-phase trials of new products.⁵

A trial of a whole food, a traditional practice, a non-patentable substance, a low-cost generic, or a non-pharmaceutical intervention almost never reaches the funding threshold the hierarchy requires. Substances and practices that produce no return on investment do not generate the evidence the framework recognises, and so they sit at the bottom of the hierarchy or fall off it entirely.

The hierarchy then performs a second move. When a question has not been studied at the top tier, the framework declares “insufficient evidence” or “no evidence of benefit.” Absence is treated as a finding. The reader is led to conclude that the unstudied intervention does not work, when what has actually been established is that no one has been willing to pay for the kind of study the framework demands.

The streetlight effect — searching for keys under the lamp because that is where the light is — is not a flaw of the framework. It describes how the framework operates by design. The hierarchy of evidence is a hierarchy of who can afford to generate evidence.

---

The Trials at the Top Are Built by the Sponsor

The architecture of the modern pharmaceutical trial is the second mechanism. The sponsor — the company that owns the product — controls the design.

The sponsor selects the primary endpoint. A trial of an antidepressant can be designed to measure a small change on a subjective rating scale at week six, rather than functional recovery at one year. A trial of a statin can be designed to measure relative risk reduction in cardiovascular events, rather than all-cause mortality. A trial of a cancer drug can be designed to measure progression-free survival — the time until the tumour grows on a scan — rather than overall survival.⁶ The endpoint determines what answer the trial is permitted to give.

The sponsor selects the comparator. A new drug compared against placebo where an effective comparator already exists tends to win. A new drug compared against an existing drug at the wrong dose, or in the wrong patient population, tends to win. The comparator becomes a design choice rather than a scientific reference.

The sponsor selects the population. Trials that approve drugs exclude the elderly, the polypharmacy patients, the pregnant, those with comorbid conditions, those with abnormal laboratory values, and those with histories of the very adverse events being assessed. The drug is then prescribed to the population that was excluded. The VIGOR trial of Vioxx excluded patients with significant cardiovascular risk; the drug was marketed and prescribed to a population dominated by such patients.⁷

The sponsor selects the duration. A trial of six weeks tells you nothing about a drug that will be taken for life. A trial of two years tells you nothing about lifetime cancer risk. Sponsors routinely halt trials early “for benefit” once a favourable interim result is reached, eliminating the longer follow-up that would have permitted assessment of late-emerging harms; the JUPITER statin trial discussed later was stopped at a median of 1.9 years rather than completing its planned four-year duration.⁸

The sponsor controls the data. Investigators at trial sites send data to the sponsor. The sponsor’s statisticians analyse it. The sponsor’s medical writers produce the manuscript. The 2017 Cochrane systematic review by Lundh and colleagues, examining 75 studies across multiple drug classes, found industry-sponsored research produces conclusions more favourable to the sponsor’s product than independently funded research of the same questions, with the effect persisting after adjustment for methodological quality.⁹ A 2003 BMJ analysis by Lexchin and colleagues found that industry-funded trials of new drugs produced results favourable to the sponsor’s product at roughly four times the rate of independently funded trials.¹⁰

The RCT does not measure efficacy. It measures what its sponsor designed it to measure.

---

What Happens to the Trials That Find Harm

The third mechanism concerns what happens to the data the sponsor would prefer not to publish.

In 2008, Erick Turner and colleagues at the Department of Veterans Affairs published an analysis in the New England Journal of Medicine of all FDA-registered trials of antidepressants conducted between 1987 and 2004 — 74 trials covering 12 drugs. The FDA records, obtained under freedom of information requests, showed 38 trials with positive results and 36 trials with negative or questionable results. Of the 38 positive trials, 37 were published. Of the 36 negative trials, 22 were not published at all, and 11 were published in a way that conveyed a positive outcome.¹¹ The published literature on antidepressants showed positive findings in 94% of trials. The actual data showed 51%.

This is publication bias as a system, not as accident. The same pattern has been documented for COX-2 inhibitors, antipsychotics, neuraminidase inhibitors, and statins.¹² Trials that find harm are buried. Trials that find benefit are amplified.

A second layer operates below publication. Ghostwriting — the practice of pharmaceutical companies producing manuscripts and recruiting academic authors to attach their names — has been documented across multiple drug classes. Internal Merck documents released in the Vioxx litigation showed company employees drafting clinical trial papers and review articles, then recruiting academic physicians to be listed as authors; in the litigation review by Ross and colleagues, the company author was frequently the first or last name on the draft before the academic name replaced it.¹³ Wyeth’s hormone replacement therapy promotion was supported by at least 26 ghostwritten papers published in the medical literature between 1998 and 2005, identified through documents released in litigation and analysed by Adriane Fugh-Berman.¹⁴

A third layer operates at the journals themselves. When a pharmaceutical company publishes a favourable trial in a major medical journal, it routinely purchases tens or hundreds of thousands of reprints of that article from the journal — reprints distributed to physicians by sales representatives as the academic credential for the product. Richard Smith, former editor of the BMJ, described medical journals in 2005 as “an extension of the marketing arm of pharmaceutical companies,” noting that reprint orders for industry-favourable studies can generate revenues sufficient to constitute a substantial share of a major journal’s income.¹⁵ The journals at the top of the EBM hierarchy depend financially on the companies whose products they evaluate.

The reader of the medical literature encounters what appears to be the considered opinion of an academic physician. The reader does not see the company writer who drafted the manuscript or the company statistician who selected the data presented.

---

“No Evidence of Harm”

The fourth mechanism is the laundering of absence into safety.

When a harm signal appears in post-marketing data, the framework processes it in stages. The harm is not yet established because no RCT has been designed to test for it. The harm cannot be established because the RCT that would test for it has not been conducted. The harm has not been confirmed because the studies that exist were not powered to detect it. The correlation between exposure and harm is not causation, because the gold-standard trial has not been performed. By the time the gold-standard trial is performed, if it ever is, the drug has been on the market for a decade and the harm is too widespread to deny.

Each stage uses the framework’s own standards to keep the product on the market. Each stage requires the evidence that the framework’s funding structure ensures will not be generated.

The phrase “no evidence of harm” does work the reader rarely notices. It does not mean studies were conducted and harm was not found. It usually means that studies sufficient to detect the harm were not conducted. The phrase converts absence into safety. The asymmetry is structural: “no evidence of benefit” is treated as a finding against an unfunded substance, while “no evidence of harm” is treated as a finding in favour of a marketed product.

The framework is unfalsifiable for the things it protects and fatal for the things it does not.

---

What Happens to the People Who Disagree

The fifth mechanism is the processing of dissent.

A clinician who observes a pattern of harm in patients and reports it is reasoning from unsystematic clinical experience — the bottom of the hierarchy. The observation is dismissed as anecdotal. A researcher who publishes findings unfavourable to a major product class is subjected to coordinated response: methodological critique, accusations of conflict of interest, retraction campaigns, ridicule in the trade press, and loss of funding. A physician who treats outside guideline-driven protocols is referred to the medical board. A patient who reports the harm is told the condition is unrelated, idiopathic, or psychological.

Peter Gøtzsche, co-founder of the Cochrane Collaboration and author of multiple Cochrane reviews unfavourable to the pharmaceutical industry, was expelled from the Cochrane governing board in 2018 after publishing critical analyses of mammography screening, psychiatric drug regulation, and antidepressant suicidality.¹⁶ The expulsion processed dissent.

John Ioannidis, professor of medicine at Stanford and one of the most cited scientists in medical literature, published “Why Most Published Research Findings Are False” in PLoS Medicine in 2005, demonstrating from within the establishment that the statistical and structural assumptions of the published evidence base were unreliable.¹⁷ The paper was met with hostility from those it implicated and quietly absorbed by those it embarrassed. The framework continued unchanged.

David Healy, a psychiatrist and historian of antidepressants whose research had exposed the suicidality signal in SSRI trial data, was offered the chair of the Mood and Anxiety Disorders Programme at the University of Toronto’s Centre for Addiction and Mental Health in 2000. After a public lecture in which he discussed Prozac-induced suicidality, the appointment was withdrawn. The Centre received substantial funding from Eli Lilly, the manufacturer of Prozac.¹⁸ Nancy Olivieri, a haematologist at the University of Toronto, reported safety concerns about deferiprone, a thalassaemia drug manufactured by Apotex, after observing harm in her clinical trial. Apotex threatened legal action; the university, which was at the time negotiating a substantial donation from Apotex, did not defend her academic freedom.¹⁹ Both cases preceded the EBM apparatus’s deployment against later dissenters; both established the template.

The framework does not produce truth and then defend it against error. The framework produces orthodoxy and then defends it against observation.

---

Case One: Vioxx

Merck submitted rofecoxib (Vioxx) to the FDA in November 1998 and received approval in May 1999.²⁰ The drug was a COX-2 inhibitor — a new class of anti-inflammatory marketed as gentler on the stomach than older drugs. Before its withdrawal, more than 80 million people worldwide had been prescribed Vioxx.²¹

The pivotal trial supporting cardiovascular and gastrointestinal claims was VIGOR (Vioxx Gastrointestinal Outcomes Research), published in the New England Journal of Medicine in November 2000.²² The trial compared Vioxx against naproxen in 8,076 rheumatoid arthritis patients. The publication reported that Vioxx caused fewer serious gastrointestinal events than naproxen. The published paper also reported that patients on Vioxx had four times the rate of myocardial infarction; further analysis presented to the FDA Arthritis Advisory Committee in February 2001 determined the rate to be fivefold.²³

The published explanation was that naproxen was protective, not that Vioxx was harmful. The cardioprotective effect of naproxen had not been established at this magnitude before VIGOR and was not established afterwards.²⁴ The interpretation served the sponsor.

Internal Merck documents released in subsequent litigation showed company awareness of the cardiovascular signal predating VIGOR’s publication. Internal communications discussed how to manage the signal; the published trial reports did not communicate what the internal analyses had described.²⁵ Merck withdrew Vioxx in September 2004 after the APPROVe trial, designed to assess Vioxx’s effects on colorectal polyps, demonstrated a doubling of cardiovascular events.²⁶ FDA epidemiologist David Graham testified before the Senate Finance Committee that Vioxx had caused an estimated 88,000 to 139,000 excess heart attacks, of which 30 to 40 percent were fatal.²⁷ The lower bound was approximately 26,000 American deaths.

The trial that approved the drug satisfied every requirement of evidence-based medicine. It was randomised. It was controlled. It was published in the most prestigious medical journal in the world. It supported guideline recommendations and reimbursement decisions. The framework that produced it found nothing wrong with it.

The harm was visible in the data. The harm was known to the sponsor. The harm was published in a form that obscured its meaning. The framework continued to recommend the drug for almost five years. When the drug was withdrawn, the framework was not.

---

Case Two: Statins

The statin literature illustrates how the framework presents data to maximise the appearance of benefit.

Statins reduce LDL cholesterol. Whether they reduce all-cause mortality in primary prevention — in patients without established cardiovascular disease — has been contested in the literature for two decades.²⁸ The framework has resolved the contest in favour of mass prescription.

The presentational device is relative risk reduction. A statin trial reports that the drug reduces cardiovascular events by some percentage. The figure is technically accurate. What it conceals is the absolute risk reduction — the actual difference in event rates between the treated group and the untreated group.

In the JUPITER trial of rosuvastatin (2008), patients on the drug had a 1.6% rate of major cardiovascular events over 1.9 years. Patients on placebo had a 2.8% rate.²⁹ The relative risk reduction was 44%. The absolute risk reduction was 1.2%. The number needed to treat — the number of patients who must take the drug for one to avoid an event — was approximately 95 over two years. The other ninety-four took the drug and received no cardiovascular benefit from it.

The framework permits the relative figure to be reported in the headlines, the abstract, the press release, the guideline recommendation, and the prescribing conversation. The absolute figure appears, if at all, in the body of the paper. The patient is told the drug reduces heart attack risk by 44%. The patient is not told the drug reduces their personal two-year risk from 2.8% to 1.6%.

Adverse effects are processed through related machinery. Many statin trials use a run-in period — patients are given the drug before randomisation, and those who experience adverse effects are excluded before the trial begins.³⁰ Trials then report low rates of muscle pain, cognitive impairment, and new-onset diabetes. Surveys of statin users in real clinical practice — without the run-in selection — find muscle symptoms in 10 to 25 percent of patients, against trial-reported rates in the low single digits.³¹

The 2013 Cochrane review of statins in primary prevention found a small mortality benefit and a non-trivial harm signal, particularly for new-onset diabetes.³² The framework’s response was not to question primary prevention prescribing. It was to expand it. The 2013 ACC/AHA cholesterol guidelines lowered the threshold for statin prescription and added an estimated 13 million Americans to the eligible population.³³

The framework approves the drug, designs the trials to maximise apparent benefit and minimise apparent harm, suppresses the inconvenient findings, expands the eligible population, and declares the resulting prescribing pattern to be evidence-based.

---

Case Three: SSRIs

The selective serotonin reuptake inhibitors entered the market in 1987 with fluoxetine (Prozac). They were marketed on the basis of a “chemical imbalance” theory of depression — the claim that depressed patients had low serotonin and the drugs corrected the deficiency.³⁴ The theory was never demonstrated. A 2022 systematic umbrella review by Joanna Moncrieff and colleagues, published in Molecular Psychiatry, concluded that the evidence does not support the hypothesis that depression is caused by reduced serotonin activity or concentrations.³⁵

The drugs were approved on the basis of trials showing small differences from placebo on rating-scale depression scores at six to eight weeks. The Turner analysis cited earlier in this essay found that the published literature substantially overstated the actual trial outcomes; when unpublished trials were included, the apparent efficacy fell substantially, with roughly half the trials having failed to demonstrate benefit.¹¹ A 2008 meta-analysis by Irving Kirsch and colleagues using FDA data found that the difference between SSRIs and placebo on depression rating scales fell below the threshold for clinical significance for all but the most severely depressed patients.³⁶

The harms followed the framework’s standard sequence.

Sexual dysfunction was acknowledged in trial reports at rates of 2 to 16 percent, well below the 50 to 70 percent rates documented in subsequent clinical surveys.³⁷ Post-SSRI sexual dysfunction — persistent sexual dysfunction continuing after discontinuation — was denied for two decades. The European Medicines Agency added a warning label in 2019.³⁸ Patients reporting the syndrome had been told for two decades it was not a recognised condition.

Suicidality in children and adolescents was visible in the trial data from the early 1990s. The published literature did not communicate the signal. A 2004 FDA review of pediatric trials confirmed it, and the FDA added a black box warning in October 2004.³⁹ A 2016 BMJ analysis by Tarang Sharma, Peter Gøtzsche and colleagues, using clinical study reports rather than published papers, found the suicidality signal in adults as well — and found systematic misclassification of suicide attempts as “emotional lability” in the original trial reports.⁴⁰

The dependence and withdrawal syndromes were denied for three decades. SSRI manufacturers and prescribing guidelines characterised the discontinuation syndrome as a mild, transient phenomenon affecting a small minority of patients. A 2019 systematic review by James Davies and John Read found 56% of patients experience withdrawal effects when attempting to discontinue, with 46% of those affected describing the experience as severe.⁴¹ The UK Royal College of Psychiatrists revised its position later that year, acknowledging that withdrawal could be severe and prolonged.⁴² The acknowledgement came thirty-two years after the first SSRI was approved.

Around one in eight American adults now takes an antidepressant.⁴³ The framework that approved them functioned exactly as designed.

---

Case Four: Bisphosphonates

Fosamax (alendronate) was approved by the FDA in 1995 as a treatment for established osteoporosis. The market was small. In 1995, the number of American women with documented osteoporosis was a fraction of what the drug’s commercial prospects required.

The framework supplied the market. In 1994, a World Health Organization study group convened in Rome, financed by the pharmaceutical industry, produced arbitrary diagnostic cutoffs based on bone mineral density measured against the average of a healthy thirty-year-old white woman.⁴⁴ Bone density between one and 2.5 standard deviations below this reference was named “osteopenia.” Bone density more than 2.5 standard deviations below was named “osteoporosis.” The categories were statistical cutoffs imposed on a normally distributed biological variable; they were not derived from outcome data on who actually fractured. Anna Tosteson, a member of the original WHO group, later said the categories had been intended as population research tools, not individual diagnoses; the chair of the meeting, John Kanis, said the same.⁴⁵ A subsequent 1999 WHO panel on osteoporosis cost analysis included eleven members, eight of them employed by pharmaceutical manufacturers.⁴⁶

Merck did the rest. The company established the Bone Measurement Institute, a nonprofit that lobbied for insurance coverage of bone density testing and underwrote the placement of scanners in physician offices across the United States.⁴⁷ In 1997, the FDA cleared a lower 5 mg dose of Fosamax specifically for women with osteopenia. Approximately thirty percent of postmenopausal women now had a “disease” requiring early intervention.⁴⁸

The pivotal trials applied the relative-risk-reduction architecture described in the statins case. The Fracture Intervention Trial reported a 47% relative reduction in hip fracture and a 52% relative reduction in radiographic vertebral fracture in women with established osteoporosis.⁴⁹ The trial enrolled high-risk women in whom even modest relative reductions translated into modest absolute differences. It did not study the population in which the drug was subsequently prescribed in greatest numbers — women with osteopenia, who carried substantially lower baseline fracture risk. The framework approved the prescribing pattern anyway.

The drug’s mechanism of action determined what would follow. Bisphosphonates concentrate in bone and disable osteoclasts, the cells responsible for clearing old bone and allowing new bone to be laid down in response to mechanical load. Blocking the clearing side of the remodeling cycle while leaving the building side unopposed produces bones that are denser by measurement and more brittle by behaviour — old bone that should have been replaced accumulates as highly mineralised, structurally compromised material. The scan reads higher. The bone breaks more easily.

The harm signal emerged after the prescribing pattern was established. Reports of atypical femoral fractures — spontaneous breaks in the shaft of the thighbone, occurring with little or no trauma — appeared in the orthopaedic literature from 2007.⁵⁰ The drug marketed to prevent fractures was producing a different category of fracture in long-term users. The FDA issued a safety warning in October 2010 but did not retract the prescribing recommendation; osteonecrosis of the jaw was added as a separate documented harm.⁵¹ A 2011 JAMA study reported significantly elevated risk of subtrochanteric and femoral shaft fractures in long-term bisphosphonate users.⁵²

The framework accepted an industry-funded definition of disease, converted asymptomatic women into patients, prescribed them a drug whose trials had been conducted in a different population, and continued recommending the drug after the iatrogenic loop became visible. Nothing in this sequence required new science. It required only the framework’s willingness to treat industry-supplied definitions as medical knowledge, industry-funded trials of one population as evidence about another, and a metric the drug improved (density on the scan) as a proxy for the outcome the drug worsened (the strength of the bone under stress).

---

What the Framework Was Protecting

The essay has examined EBM on its own terms. By its own stated standards, the framework does not produce reliable knowledge, does not protect patients, does not exclude bias, and does not distinguish truth from manufactured evidence. Every claim it makes for itself can be inverted with examples from its own literature.

A deeper question remains.

EBM is the epistemological enforcement layer for a particular model of medicine. That model holds that the body is a malfunctioning machine, that illness is caused by external invaders or internal genetic defects, that the response to illness is the introduction of a patented chemical or biological product, that the practitioner’s task is to match product to diagnosis, and that the evidence required to justify the product is the kind of trial pharmaceutical companies are equipped to manufacture.

The framework’s structural bias against whole foods, traditional practices, low-cost interventions, and reasoning from mechanism is not incidental to this model. The pharmaceutical paradigm requires the bias to function. A framework that recognised the body’s intrinsic capacity for self-regulation and repair, that attended to the actual sources of damage — toxic exposure, nutritional depletion, electromagnetic burden, psychological strain — and that treated the practitioner’s task as the removal of obstacles rather than the introduction of products, would not need RCTs at the top of its hierarchy because it would not be generating products to be tested. It would be observing what the body does when the obstacles are removed.

EBM is not science being corrupted by industry. EBM is the epistemological form industry required. It was designed in the 1990s, codified in the 2000s, and operated at civilisational scale in the 2020s. The framework has done what its architecture predicted.

This is what political economist Toby Rogers, in testimony before the United States Senate in 2025, named epistemic capture — the colonisation of knowledge production itself rather than the regulation of its products.⁵³ When an industry captures regulation, it controls decisions. When an industry captures epistemology, it controls what is allowed to count as a fact. EBM is epistemic capture’s operating system in medicine. The same investment funds that hold major positions in the pharmaceutical companies whose products the framework evaluates also hold major positions in the publishers of the journals that perform the evaluation; the producer and the certifier of medical knowledge share their owners.⁵⁴

The clinician at the bedside who notices a pattern across patients and adjusts their practice accordingly is doing what physicians have done for thousands of years. The framework calls this anecdote and ranks it at the bottom of the hierarchy. The patient who recovers from a chronic condition through dietary change, sunlight, sleep, and the removal of pharmaceutical exposures is doing something the framework cannot measure, because the framework was not designed to measure it. The traditional practitioner whose people have used a plant for fifteen generations is reasoning from a form of evidence the framework excludes by definition.

What the framework calls evidence is what industry can pay for. What the framework calls anecdote is what the body actually does.

The 1992 paper announced a paradigm shift. The shift was real. The direction was not what the paper claimed. Medicine did not move from intuition to evidence. It moved from the physician’s accumulated judgement to the industry’s manufactured documentation. The hierarchy of evidence is the hierarchy of who paid for the study. Evidence-based medicine, examined as machinery, is evidence-biased medicine — biased structurally, by what gets funded; biased methodologically, by what gets measured; biased editorially, by what gets published; biased financially, by who owns the journals; biased institutionally, by what gets recommended after the harm has been documented. The bias is the framework.

The document is publicly available. The signatures are on it. The date is November 1992. Whatever the framework was sold as, the framework is what its architecture produced. What its architecture produced is in the medical journals, in the prescribing patterns, in the disability statistics, and in the cemeteries. It is also in the minds of two generations of doctors and patients who no longer believe their own observations count.

---

References

1. Evidence-Based Medicine Working Group. “Evidence-Based Medicine: A New Approach to Teaching the Practice of Medicine.” JAMA 268, no. 17 (November 1992): 2420–2425.

2. Ferguson J. “BMJ Readers Choose the ‘Sanitary Revolution’ as Greatest Medical Advance Since 1840.” BMJ 334, no. 7585 (January 2007): 111.

3. OCEBM Levels of Evidence Working Group. “The Oxford 2011 Levels of Evidence.” Oxford Centre for Evidence-Based Medicine.

4. Sertkaya A, Wong HH, Jessup A, Beleche T. “Key Cost Drivers of Pharmaceutical Clinical Trials in the United States.” Clinical Trials 13, no. 2 (April 2016): 117–126.

5. Ehrhardt S, Appel LJ, Meinert CL. “Trends in National Institutes of Health Funding for Clinical Trials Registered in ClinicalTrials.gov.” JAMA 314, no. 23 (December 2015): 2566–2567.

6. Kim C, Prasad V. “Cancer Drugs Approved on the Basis of a Surrogate End Point and Subsequent Overall Survival.” JAMA Internal Medicine 175, no. 12 (December 2015): 1992–1994.

7. Topol EJ. “Failing the Public Health: Rofecoxib, Merck, and the FDA.” New England Journal of Medicine 351, no. 17 (October 2004): 1707–1709.

8. Bassler D, Briel M, Montori VM, et al. “Stopping Randomized Trials Early for Benefit and Estimation of Treatment Effects: Systematic Review and Meta-Regression Analysis.” JAMA 303, no. 12 (March 2010): 1180–1187.

9. Lundh A, Lexchin J, Mintzes B, Schroll JB, Bero L. “Industry Sponsorship and Research Outcome.” Cochrane Database of Systematic Reviews, no. 2 (February 2017): MR000033.

10. Lexchin J, Bero LA, Djulbegovic B, Clark O. “Pharmaceutical Industry Sponsorship and Research Outcome and Quality: Systematic Review.” BMJ 326, no. 7400 (May 2003): 1167–1170.

11. Turner EH, Matthews AM, Linardatos E, Tell RA, Rosenthal R. “Selective Publication of Antidepressant Trials and Its Influence on Apparent Efficacy.” New England Journal of Medicine 358, no. 3 (January 2008): 252–260.

12. Jefferson T, Jones MA, Doshi P, et al. “Neuraminidase Inhibitors for Preventing and Treating Influenza in Adults and Children.” Cochrane Database of Systematic Reviews, no. 4 (April 2014): CD008965.

13. Ross JS, Hill KP, Egilman DS, Krumholz HM. “Guest Authorship and Ghostwriting in Publications Related to Rofecoxib: A Case Study of Industry Documents from Rofecoxib Litigation.” JAMA 299, no. 15 (April 2008): 1800–1812.

14. Fugh-Berman AJ. “The Haunting of Medical Journals: How Ghostwriting Sold ‘HRT’.” PLoS Medicine 7, no. 9 (September 2010): e1000335.

15. Smith R. “Medical Journals Are an Extension of the Marketing Arm of Pharmaceutical Companies.” PLoS Medicine 2, no. 5 (May 2005): e138.

16. Gøtzsche PC. Death of a Whistleblower and Cochrane’s Moral Collapse. People’s Press, 2019.

17. Ioannidis JPA. “Why Most Published Research Findings Are False.” PLoS Medicine 2, no. 8 (August 2005): e124.

18. Healy D. Let Them Eat Prozac: The Unhealthy Relationship Between the Pharmaceutical Industry and Depression. New York University Press, 2004. See also Healy D. “Conflicting Interests in Toronto: Anatomy of a Controversy at the Interface of Academia and Industry.” Perspectives in Biology and Medicine 45, no. 2 (Spring 2002): 250–263.

19. Thompson J, Baird P, Downie J. The Olivieri Report: The Complete Text of the Report of the Independent Inquiry Commissioned by the Canadian Association of University Teachers. James Lorimer & Company, 2001. See also Schafer A. “Biomedical Conflicts of Interest: A Defence of the Sequestration Thesis—Learning from the Cases of Nancy Olivieri and David Healy.” Journal of Medical Ethics 30, no. 1 (February 2004): 8–24.

20. US Food and Drug Administration. Approval Letter for Vioxx (Rofecoxib), May 20, 1999.

21. Krumholz HM, Ross JS, Presler AH, Egilman DS. “What Have We Learnt from Vioxx?” BMJ 334, no. 7585 (January 2007): 120–123.

22. Bombardier C, Laine L, Reicin A, et al. “Comparison of Upper Gastrointestinal Toxicity of Rofecoxib and Naproxen in Patients with Rheumatoid Arthritis.” New England Journal of Medicine 343, no. 21 (November 2000): 1520–1528.

23. Curfman GD, Morrissey S, Drazen JM. “Expression of Concern: Bombardier et al., ‘Comparison of Upper Gastrointestinal Toxicity of Rofecoxib and Naproxen.’” New England Journal of Medicine 353, no. 26 (December 2005): 2813–2814.

24. Jüni P, Nartey L, Reichenbach S, Sterchi R, Dieppe PA, Egger M. “Risk of Cardiovascular Events and Rofecoxib: Cumulative Meta-Analysis.” The Lancet 364, no. 9450 (December 2004): 2021–2029.

25. Berenson A. “Evidence in Vioxx Suits Shows Intervention by Merck Officials.” New York Times, April 24, 2005.

26. Bresalier RS, Sandler RS, Quan H, et al. “Cardiovascular Events Associated with Rofecoxib in a Colorectal Adenoma Chemoprevention Trial.” New England Journal of Medicine 352, no. 11 (March 2005): 1092–1102.

27. Graham DJ. Testimony before the US Senate Finance Committee, November 18, 2004.

28. Abramson J, Wright JM. “Are Lipid-Lowering Guidelines Evidence-Based?” The Lancet 369, no. 9557 (January 2007): 168–169.

29. Ridker PM, Danielson E, Fonseca FA, et al. “Rosuvastatin to Prevent Vascular Events in Men and Women with Elevated C-Reactive Protein.” New England Journal of Medicine 359, no. 21 (November 2008): 2195–2207.

30. Newman CB, Preiss D, Tobert JA, et al. “Statin Safety and Associated Adverse Events: A Scientific Statement from the American Heart Association.” Arteriosclerosis, Thrombosis, and Vascular Biology 39, no. 2 (February 2019): e38–e81.

31. Buettner C, Davis RB, Leveille SG, Mittleman MA, Mukamal KJ. “Prevalence of Musculoskeletal Pain and Statin Use.” Journal of General Internal Medicine 23, no. 8 (August 2008): 1182–1186.

32. Taylor F, Huffman MD, Macedo AF, et al. “Statins for the Primary Prevention of Cardiovascular Disease.” Cochrane Database of Systematic Reviews, no. 1 (January 2013): CD004816.

33. Pencina MJ, Navar-Boggan AM, D’Agostino RB Sr, et al. “Application of New Cholesterol Guidelines to a Population-Based Sample.” New England Journal of Medicine 370, no. 15 (April 2014): 1422–1431.

34. Lacasse JR, Leo J. “Serotonin and Depression: A Disconnect between the Advertisements and the Scientific Literature.” PLoS Medicine 2, no. 12 (December 2005): e392.

35. Moncrieff J, Cooper RE, Stockmann T, Amendola S, Hengartner MP, Horowitz MA. “The Serotonin Theory of Depression: A Systematic Umbrella Review of the Evidence.” Molecular Psychiatry 28 (July 2022): 3243–3256.

36. Kirsch I, Deacon BJ, Huedo-Medina TB, et al. “Initial Severity and Antidepressant Benefits: A Meta-Analysis of Data Submitted to the Food and Drug Administration.” PLoS Medicine 5, no. 2 (February 2008): e45.

37. Montejo AL, Llorca G, Izquierdo JA, Rico-Villademoros F. “Incidence of Sexual Dysfunction Associated with Antidepressant Agents.” Journal of Clinical Psychiatry 62, suppl. 3 (2001): 10–21.

38. European Medicines Agency. PRAC Recommendations on Signals, May 13–16, 2019.

39. US Food and Drug Administration. “FDA Public Health Advisory: Suicidality in Children and Adolescents Being Treated with Antidepressant Medications.” October 15, 2004.

40. Sharma T, Guski LS, Freund N, Gøtzsche PC. “Suicidality and Aggression During Antidepressant Treatment: Systematic Review and Meta-Analyses Based on Clinical Study Reports.” BMJ 352 (January 2016): i65.

41. Davies J, Read J. “A Systematic Review into the Incidence, Severity and Duration of Antidepressant Withdrawal Effects.” Addictive Behaviors 97 (October 2019): 111–121.

42. Royal College of Psychiatrists. “Position Statement on Antidepressants and Depression.” PS04/19, May 2019.

43. Brody DJ, Gu Q. “Antidepressant Use Among Adults: United States, 2015–2018.” NCHS Data Brief No. 377. National Center for Health Statistics, September 2020.

44. Petersen M. “Hormones and Bones: How the Drug Companies Made Aging Into a Disease.” New York Times, November 16, 2003. See also Heaney RP et al. and consensus development conference: “Diagnosis, Prophylaxis, and Treatment of Osteoporosis.” American Journal of Medicine 94, no. 6 (June 1993): 646–650, and Kanis JA, WHO Study Group. “Assessment of Fracture Risk and Its Application to Screening for Postmenopausal Osteoporosis: Report of a WHO Study Group.” WHO Technical Report Series 843. Geneva: World Health Organization, 1994.

45. Spiegel A. “How a Bone Disease Grew to Fit the Prescription.” NPR Morning Edition, December 21, 2009.

46. Alonso-Coello P, García-Franco AL, Guyatt G, Moynihan R. “Drugs for Pre-Osteoporosis: Prevention or Disease Mongering?” BMJ 336, no. 7636 (January 2008): 126–129.

47. Brownlee S. Overtreated: Why Too Much Medicine Is Making Us Sicker and Poorer. Bloomsbury, 2007.

48. Moynihan R. “Osteoporosis and the Marketing of a Disease.” Therapeutics Initiative Workshop, University of British Columbia, October 2006. Fosamax sales figures: Merck Annual Reports, 1996–2005.

49. Black DM, Cummings SR, Karpf DB, et al. “Randomised Trial of Effect of Alendronate on Risk of Fracture in Women with Existing Vertebral Fractures.” The Lancet 348, no. 9041 (December 1996): 1535–1541. See also Black DM, Thompson DE, Bauer DC, et al. “Fracture Risk Reduction with Alendronate in Women with Osteoporosis: The Fracture Intervention Trial.” Journal of Clinical Endocrinology & Metabolism 85, no. 11 (November 2000): 4118–4124.

50. Neviaser AS, Lane JM, Lenart BA, Edobor-Osula F, Lorich DG. “Low-Energy Femoral Shaft Fractures Associated with Alendronate Use.” Journal of Orthopaedic Trauma 22, no. 5 (May 2008): 346–350. Earlier case reports beginning 2005; pattern recognised in the literature from 2007.

51. US Food and Drug Administration. “Safety Update for Osteoporosis Drugs, Bisphosphonates, and Atypical Fractures.” FDA Drug Safety Communication, October 13, 2010.

52. Park-Wyllie LY, Mamdani MM, Juurlink DN, et al. “Bisphosphonate Use and the Risk of Subtrochanteric or Femoral Shaft Fractures in Older Women.” JAMA 305, no. 8 (February 2011): 783–789. See also Dell RM, Adams AL, Greene DF, et al. “Incidence of Atypical Nontraumatic Diaphyseal Fractures of the Femur.” Journal of Bone and Mineral Research 27, no. 12 (December 2012): 2544–2550.

53. Rogers T. Testimony before the United States Senate Committee on Homeland Security and Governmental Affairs, September 10, 2025. See also Rogers T. “Big Pharma’s Epistemic Capture of Medical Research.” uTobian (Substack), September 2025.

54. Bazin X. Big Pharma démasqué: Médicaments dangereux, vaccins controversés... Quand l’industrie pharmaceutique nous prend pour des cobayes. Trédaniel, 2022. Common ownership structures documented through RELX Group annual reports and SEC filings of major institutional shareholders (BlackRock, Vanguard, State Street) in both pharmaceutical companies and scientific publishing groups (RELX, Springer Nature, Wiley).