Iatrogenics

Drilling holes into the skull to cure headaches, epilepsy, and mental illness — one of the oldest surgical procedures ever documented. Trephined skulls have been found across Europe, Asia, and the Americas. Some patients survived, evidenced by healed bone growth, but the procedure's rationale was essentially magical.

"First, do no harm" — the recognition that intervention carries risk is as old as medicine itself. Yet Hippocrates also codified humoral theory, which justified bloodletting for the next two millennia.

For roughly 2,000 years, draining patients' blood was standard treatment for nearly everything — sore throats, fevers, plague, childbirth complications. Based on the theory that illness resulted from humoral imbalance. Lancets, fleams, and leeches were a physician's primary tools. Marie-Antoinette was bled during labor; Queen Anne died two days after her doctors arrived.

Mercury — as liquid quicksilver, calomel pills, ointments, and vapor baths — was prescribed for syphilis, constipation, melancholy, parasites, and influenza for four centuries. Patients lost teeth, developed tremors, cognitive decline, and gangrenous facial holes. Chinese Emperor Qin Shi Huang reportedly died from mercury pills designed to make him immortal. The treatment persisted until antibiotics made it obsolete.

When the king suffered a fit while shaving, fourteen physicians descended. Over five days they bled him, blistered him, gave emetics, shaved and blistered his head, administered enemas, applied plasters of pigeon droppings, dosed him with calomel and antimony, and bled him again. He died — almost certainly more from the treatment than from whatever originally caused the fit. A distilled case of iatrogenic overkill by committee.

Benjamin Rush — the "American Hippocrates" — championed extreme bloodletting and calomel purging as standard practice. During Philadelphia's 1793 yellow fever epidemic, Rush drained patients of massive blood volumes and dosed them with mercury to the point of acute poisoning. His students spread these methods across the country. Patients routinely died not from their disease but from the treatment.

Physicians drained an estimated 5–9 pints of blood in under 16 hours, administered calomel and cantharidin blisters to treat a throat infection. Washington died the next day. A textbook case: the treatment killed faster than the disease could have.

Ignaz Semmelweis proved that doctors' unwashed hands transmitted fatal puerperal fever between patients in Vienna's maternity wards. Mortality dropped from ~18% to ~2% with handwashing. The medical establishment rejected his findings and ridiculed him. He was committed to an asylum, where he died at 47 — of a wound infection, the very condition he'd spent his life fighting.

Shot by an assassin at a Washington D.C. train station, the bullet lodged behind Garfield's pancreas but hit no vital organs — modern consensus holds he likely would have survived. Instead, twelve different doctors probed the wound with unsterilized fingers and instruments, beginning on the germ-infested station floor. Lead physician Dr. D. Willard Bliss (whose first name was literally "Doctor") rejected Lister's antiseptic techniques, repeatedly plunging unsterilized probes into the wound over 80 agonizing days, even puncturing the president's liver. He refused to let Alexander Graham Bell scan both sides of the body with an experimental metal detector, insisting the bullet was on the right (the autopsy found it on the left). Garfield wasted from 210 to 130 pounds, riddled with pus-filled abscesses, fed rectally, and dosed with opium. He died of sepsis on September 19. His assassin's defense at trial: "I shot the president, but his physicians killed him."

António Egas Moniz invented the prefrontal lobotomy in 1935 and received the Nobel Prize for it in 1949. Walter Freeman popularized a transorbital version in the U.S., driving across the country in his "lobotomobile" performing the procedure with an ice pick through the eye socket on unanesthetized patients. Roughly one-third of patients didn't survive. Those who did often suffered permanent cognitive impairment, emotional blunting, and personality destruction. The procedure wasn't abandoned until psychiatric drugs emerged in the 1960s.

Batches of Jonas Salk's inactivated polio vaccine produced by Cutter Laboratories contained live virus instead of killed virus. About 120,000 children received the defective vaccines; roughly 40,000 developed abortive polio, 56 were paralyzed, and 5 died. The incident devastated public confidence in the Salk vaccine and led directly to federal regulation of vaccine manufacturing — but not before the damage was done.

An estimated 10–30% of polio vaccines administered in the U.S. were contaminated with simian virus 40 (SV40), a monkey virus from the kidney cell cultures used to produce the vaccine. Roughly 90% of children and 60% of adults in the U.S. received potentially contaminated doses. SV40 causes cancer in laboratory animals, though epidemiological studies have not conclusively established a cancer link in humans. The contamination continued for eight years before testing requirements were established.

Marketed as a safe sedative for morning sickness, thalidomide caused severe birth defects — shortened or absent limbs — in over 10,000 children across 46 countries. Triggered the "third wave" of iatrogenic suspicion and led directly to modern drug regulation frameworks, including the 1962 Kefauver-Harris Amendment requiring proof of efficacy before approval.

A synthetic estrogen prescribed to millions of pregnant women to prevent miscarriage — despite a 1953 randomized trial showing it didn't work. Decades later, their daughters developed rare vaginal cancers and reproductive abnormalities. A landmark case of iatrogenic harm crossing generations.

Ivan Illich published the seminal critique arguing that industrial medicine had become a major threat to health. Introduced the clinical/social/cultural framework for understanding iatrogenesis — shifting the analysis from individual error to systemic pathology.

The King of Rock and Roll died at 42 from cardiac arrhythmia aggravated by a cocktail of prescription drugs — barbiturates, codeine, and sedatives. His personal physician, Dr. George Nichopoulos, had prescribed over 10,000 doses of various drugs to Presley in the eight months before his death. His defense: "I cared too much." His medical license was eventually revoked for a pattern of overprescription. Taleb would later cite personal physicians as particularly vulnerable to naive interventionism — they need to justify their salaries and prove they're "doing something."

A mass vaccination campaign against a feared swine flu pandemic immunized over 40 million Americans. The pandemic never materialized, but the vaccine caused an increased risk of Guillain-Barré Syndrome (GBS) — approximately 1 additional case per 100,000 vaccinated. The program was halted. The IOM later confirmed the causal link. A textbook case of iatrogenics from premature large-scale intervention under uncertainty: the cure deployed against a threat that didn't arrive.

The first licensed rotavirus vaccine was pulled from the market after reports that it caused intussusception — a rare, dangerous bowel obstruction where the intestine folds into itself like a telescope — in infants. The risk was identified through post-licensure surveillance, highlighting a recurring pattern: pre-approval trials too small to detect rare but real adverse events that only surface after millions of doses are administered.

The Institute of Medicine estimated 44,000–98,000 annual deaths from medical errors in U.S. hospitals, launching patient safety as a formal field. The report's central argument: most errors are systems problems, not individual failures.

Merck's blockbuster painkiller, prescribed to 80 million people worldwide, was withdrawn after studies linked it to increased heart attacks and strokes. Estimates suggest it may have caused between 88,000–140,000 cases of serious heart disease in the U.S. alone. Internal documents later revealed the company had been aware of cardiovascular risks for years.

Purdue Pharma's aggressive marketing of OxyContin, backed by claims that addiction risk was minimal, triggered the worst drug epidemic in American history. Over 2 million Americans developed opioid use disorders from prescribed medications. Systematic reviews show roughly 3% incidence of iatrogenic opioid addiction among chronic pain patients — with rates climbing to 6% at high doses over long durations. The crisis has killed over 500,000 Americans since 1999.

Jackson died at 50 from acute propofol intoxication — a surgical anesthetic his personal physician, Dr. Conrad Murray, had been administering nightly to treat insomnia. Using propofol for sleep is a radical misapplication of a drug designed for operating rooms. Murray was convicted of involuntary manslaughter. The case crystallized the "agency problem" in iatrogenics: Murray was being paid $150,000/month by Jackson's concert promoter and had powerful financial incentives to comply with his patient's demands rather than exercise medical judgment.

Generalized iatrogenics beyond medicine to all complex systems. Introduced naive interventionism, the agency problem, and via negativa as structural frameworks. Key insight: we systematically over-intervene where benefits are marginal and under-intervene in genuine emergencies.

The 72nd World Health Assembly designated September 17 as World Patient Safety Day, formally acknowledging iatrogenic harm as a global public health priority requiring systemic — not just individual — responses.

Stimulant prescriptions in the U.S. nearly doubled from 50.5 million (2012) to over 80 million (2022), while ADHD prevalence rose from 6.1% to 10.2% in children (1997–2016). Amphetamine production tripled from ~8 million grams (2006) to over 25 million grams (2021). The U.S. accounts for less than 5% of the world's population but 83% of global ADHD medication volume. Former APA president Jeffrey Lieberman noted: "The problem is not so much that we have a shortage of medication, but an overdiagnosis of the condition." Telehealth companies like Cerebral and Done Global faced federal investigations for operating as "pill mills" — the CEO of Done Global was indicted in 2024. Stimulant prescriptions now trend similarly to the trajectory of opioids in the early 2000s. Illich's social iatrogenesis made manifest: the medicalization of distraction, performance pressure, and normal human variation, amplified by social media self-diagnosis and a healthcare system with financial incentives to prescribe.

Early pandemic guidelines recommended "very early" intubation for COVID patients requiring supplemental oxygen above 6 L/min, driven by fears of viral aerosolization and rapid respiratory decline. A widely-cited New York study initially reported 88% mortality for ventilated patients (later corrected to ~25% once incomplete data was accounted for, but the damage to clinical thinking was done). A meta-analysis of 57,420 patients across 69 studies found overall ventilator mortality of ~45%. Researchers later recognized many COVID patients presented with atypical ARDS — profoundly low oxygen but still compliant, flexible lungs — making mechanical ventilation a poor fit. As one PMC review noted, "the ventilator was a source of iatrogenic injury" and the imperative to intubate "subtly shaped and constrained medical thinking." The field eventually shifted toward high-flow nasal oxygen and delayed intubation, but an unknown number of patients were intubated unnecessarily during the critical early months.

The rapid development and mass deployment of COVID-19 vaccines — which saved millions of lives globally — also produced documented iatrogenic harms at population scale. The National Academies of Sciences (2024) established a causal relationship between mRNA vaccines (Pfizer, Moderna) and myocarditis, with incidence peaking in young males at ~1 in 16,750 after the second dose (1 in 32,000 overall). The FDA required updated warning labels after a study found persistent cardiac MRI abnormalities at 5-month follow-up. Adenovirus-vector vaccines (AstraZeneca, J&J) caused vaccine-induced immune thrombocytopenia and thrombosis (VITT) — rare but potentially fatal blood clotting — at rates of ~3 per 100,000 doses (AstraZeneca, UK data) and ~4 per million (J&J, U.S. data). AstraZeneca withdrew its vaccine globally in May 2024 and acknowledged the TTS link in ongoing litigation. A 2026 NEJM study identified the specific genetic mutation (IGLV3-21*02) predisposing carriers to VITT. The episode illustrates the irreducible tension in mass public health interventions: aggregate population benefit coexisting with individual iatrogenic harm — exactly the asymmetry Taleb describes when intervening in complex systems under uncertainty.

The STOP IATRO project in Spain, Choosing Wisely campaigns internationally, and a 2026 scoping review finding overdiagnosis in nearly every clinical field signal growing institutional recognition. The question shifting in medical culture: from "What can we do?" to "What should we refrain from doing?"

By mid-2024, roughly 950 AI/ML-enabled medical devices had received FDA clearance, with ~100 new approvals each year and a market projected to grow from $13.7 billion (2024) to over $255 billion by 2033. A 2024 survey found 66% of physicians already using AI in their practice, up from 38% in 2023. ECRI — the global healthcare safety nonprofit — named AI the #1 health technology hazard for 2025. Oxford ethicists have proposed the term "AI-trogenic harm" to describe a new category of iatrogenesis where, unlike traditional iatrogenic harm that maps to a discrete clinical act by a known individual, responsibility disperses across developers, institutions, datasets, and opaque algorithms — making attribution of harm nearly impossible.

The risks map precisely onto Taleb's four conditions for iatrogenics: complexity beyond comprehension (deep learning models whose reasoning is opaque even to their creators — "black box" medicine at industrial scale); separation from consequences (developers and deployers of AI tools are structurally insulated from the patients harmed by their outputs); action bias (AI systems are built to generate outputs, never to recommend abstention — they have no concept of via negativa); and the agency problem (financial incentives push rapid deployment while the costs of error are borne by patients). AI hallucinations — confidently stated falsehoods — have already generated false diagnoses and treatment plans. Training data biases systematically underserve populations underrepresented in datasets, producing a form of structural clinical iatrogenesis. Insurance companies are using AI algorithms to override treating physicians' medical necessity determinations, and the DOJ in 2024 subpoenaed healthcare companies over whether generative AI in electronic medical records was generating excessive or medically unnecessary care.

Perhaps most critically, AI threatens to accelerate Illich's cultural iatrogenesis — the destruction of the capacity for autonomous coping — by further displacing clinical judgment with algorithmic authority. When a physician's reasoning is overridden by a model they cannot interrogate, the system doesn't just risk individual errors: it systematically degrades the epistemic infrastructure of medicine itself. The question is not whether AI will produce iatrogenic harm — it already has. The question is whether the same institutional dynamics that enabled opioids, ventilator overuse, and amphetamine overprescription will constrain AI's integration into medicine, or whether the cycle will repeat at computational speed.