Mechanism #14: Unfalsifiable Entrenchment — when a hypothesis becomes impossible to disprove because it mutates after every failure while accumulating too much institutional investment to abandon.
"I hate the phrase, 'the amyloid hypothesis.' It has been called the amyloid plaque hypothesis, the amyloid-beta hypothesis, the amyloid cascade hypothesis, the amyloid oligomer hypothesis. Because it continually changes, you can never disprove it."
— Thomas Südhof, Nobel laureate in Physiology or Medicine, February 2025
For thirty years, the dominant theory of Alzheimer's disease has been that amyloid-beta protein accumulates into plaques that destroy neurons and cause dementia. Remove the plaques, cure the disease. The theory is elegant. The institutional investment is staggering — $3.8 billion per year in NIH Alzheimer's funding, roughly half on amyloid-related projects. Tens of billions more from pharmaceutical companies. Thousands of careers. An entire FDA regulatory pathway.
The results: 98 failed drug compounds. A 99.6% clinical trial failure rate. Between 150 and 300 experimental treatments that showed no clinical benefit. And a hypothesis that absorbed every single failure by shape-shifting into a new version of itself.
The Mutations
A falsifiable hypothesis dies when the evidence contradicts it. An unfalsifiable one mutates. Here is what three decades of mutation look like:
Each failure generates a new modification. Each modification preserves the hypothesis. The goalposts don't just move — they dissolve and reform somewhere new. A hypothesis that has been the plaques, then the oligomers, then the timing, then combination therapy, is not a hypothesis. It is a research program that has immunized itself against evidence.
The Paradox at the Heart
The sharpest version of this contradiction comes from the drugs themselves.
Stanford neurologist Michael Greicius: "All are quite adept at pulling amyloid plaques out of the brain. But none of these drugs makes much of a difference to the patient's wellbeing." His research shows that tau tangles — not amyloid — correlate with the brain regions that are actually damaged. The amyloid PET scans look the same regardless of symptoms.
Even Dennis Selkoe, one of the original champions of the amyloid hypothesis, gave a remarkably candid assessment when asked if anti-amyloid drugs would ever improve cognition or cure the disease: "I doubt it."
Meanwhile, the cost of trying. Donanemab's trial: 24% of participants developed brain swelling (ARIA), 31% brain bleeding. Three deaths attributed to the drug. Nineteen total deaths in the trial arm. Lecanemab costs $26,500 per year — and quadruples with required monitoring scans. Estimated Medicare burden: $2–5 billion annually.
Why It Can't Stop
This is where unfalsifiable entrenchment separates from mere stubbornness. The hypothesis persists not just because people believe in it but because the infrastructure requires it.
NIH allocated $3.8 billion to Alzheimer's research in 2025 — a sevenfold increase since 2011. Roughly half flows to amyloid-related projects. Alternative researchers report being systematically sidelined by what some call the "amyloid mafia." Junior scientists who question the hypothesis risk their careers. Grant review panels are populated by amyloid researchers who evaluate amyloid proposals.
Pharmaceutical companies have sunk tens of billions into the pipeline. Eisai's stock moves on lecanemab data. Biogen's entire Alzheimer's strategy was aducanumab. The FDA built an accelerated approval pathway specifically for amyloid clearance as a surrogate endpoint — the exact framework that the advisory committee was told would not be used, until it was.
Südhof sees the approved drugs themselves as a trap: "I am not sure that the approval of the antibodies is actually a positive thing. They might actually slow down the field by providing a commercial impediment to developing advances that might provide a bigger impact."
This is Goodhart's Law at civilizational scale. The measure (amyloid clearance) became the target and ceased to be a good measure of what it was supposed to indicate (cognitive improvement). The entire system — funding, careers, regulatory pathways, pharmaceutical pipelines — is now optimized for a proxy that doesn't predict the outcome.
What Amyloid Actually Is
Emerging research tells a different story. Amyloid may be a symptom, not a cause — possibly even a protective response. Osaka University researchers found in November 2025 that the brain's waste clearance system becomes impaired early in the disease, and amyloid accumulates as a downstream consequence. A March 2026 paper in PNAS Nexus supports the hypothesis that autophagy failure acts upstream of both amyloid and tau pathology.
The field is slowly pivoting. Over 60% of the 2025 Alzheimer's pipeline now targets mechanisms beyond amyloid — tau, neuroinflammation, vascular dysfunction, metabolic pathways. More than 50 inflammation-related agents are in clinical trials. The Frontiers in Pharmacology review from 2025 calls for a fundamental re-evaluation of the amyloid framework.
But the pivot is agonizingly slow. Thirty years of infrastructure doesn't redirect overnight. The amyloid hypothesis still consumes the largest share of funding, still shapes the regulatory pathway, still dominates grant review.
Mechanism #14
This is different from the thirteen mechanisms I've mapped before. Definition manipulation (#1) redefines success until the product qualifies. Narrative maintenance (#7) keeps a story alive because institutions need it. Regulatory surrogate (#10) substitutes an approval standard for a clinical utility standard.
Unfalsifiable entrenchment combines all of these — and adds the critical element that makes it uniquely destructive: the hypothesis itself evolves. It's not just that the system resists correction. The idea at the center of the system shapeshifts to absorb every correction. Plaques → oligomers → timing → combination → whatever comes next.
A hypothesis that can explain any result explains nothing. A research enterprise organized around an unfalsifiable idea will never find a cure — it will find increasingly sophisticated reasons why the cure hasn't worked yet.
Meanwhile, 6.9 million Americans live with Alzheimer's. One in three seniors dies with dementia. The disease killed more people than breast cancer and prostate cancer combined in 2024. And the dominant research paradigm just spent thirty years and tens of billions of dollars perfecting the removal of a protein that may not cause the disease.
Sources: STAT News / Charles Piller · Stanford / Greicius · Science / Schrag · NPR / FDA resignations · PMC / trial failures · Nature / NICE rejection · Frontiers / amyloid re-evaluation