The Alzheimer's Science Problem: Why the Amyloid Hypothesis Keeps Failing Drug Trials

Here's the weird part about Alzheimer's disease research: we've had a leading hypothesis for over 30 years, we've spent billions testing drugs based on that hypothesis, those drugs have failed repeatedly, and we've mostly continued pursuing the same hypothesis anyway. That's either admirable scientific persistence or sunk-cost thinking at an institutional scale. Probably some of both.

## What the amyloid hypothesis actually claims

In 1984, molecular biologist Glenner and Wong identified that the amyloid plaques found in the brains of Alzheimer's patients were composed of a specific peptide — beta-amyloid, cleaved from a larger protein called APP (amyloid precursor protein). The amyloid hypothesis, formally articulated by John Hardy and David Allsop in 1991, proposed that the accumulation of these beta-amyloid plaques drives the disease cascade: plaques form, trigger tau tangles (another protein that accumulates abnormally in Alzheimer's), neurons die, and cognitive function degrades.

The genetic evidence was compelling. Mutations in APP and in the PSEN1 and PSEN2 genes (which affect amyloid processing) cause rare familial forms of Alzheimer's. People with Down syndrome, who carry an extra copy of chromosome 21 (where APP is encoded), develop Alzheimer's pathology earlier. *The circumstantial case for amyloid looked strong.*

## Why the trials kept failing

The failure rate has been spectacular. Between 2002 and 2012, roughly 99.6% of Alzheimer's clinical trials failed — a failure rate higher than almost any other disease category. Many of those were amyloid-targeting drugs.

The pattern went roughly like this: drug company develops antibody or small molecule that clears amyloid from the brain, runs Phase 3 trial, drug successfully clears amyloid on PET scan, cognitive decline continues anyway. Patients treated with bapineuzumab, solanezumab, crenezumab — amyloid levels dropped. Cognition didn't improve.

The most charitable interpretation: we were treating patients too late. By the time clinical symptoms appear, the amyloid accumulation has been occurring for 10 to 20 years. Neurons already lost don't return. Maybe the hypothesis is correct, but the intervention window is earlier than we've been testing in.

This motivated the A4 Study (Anti-Amyloid Treatment in Asymptomatic Alzheimer's) — treating cognitively normal adults who showed elevated amyloid on PET scans before symptoms appeared. Results published in 2023 showed solanezumab failed to prevent cognitive decline even in this population. That was a significant blow to the treatment-timing explanation.

> 🔬 **Quick experiment:** Look up the timeline of amyloid PET imaging availability (available since roughly 2004) and map it against when major trial failures were announced. Notice that many trials were designed before we could even confirm amyloid burden at enrollment — participants were often selected on symptoms and age alone, meaning some proportion in these trials may not have even had the amyloid pathology the drug was targeting.

## What the data fraud scandal changed

In 2022, a Science investigation found that Sylvain Lesné, a prominent researcher at the University of Minnesota, had manipulated images in papers reporting a key amyloid oligomer finding — specifically, evidence for a particular form of beta-amyloid called Aβ*56 as causative in Alzheimer's. These papers had been widely cited and had influenced the direction of subsequent research.

The fraud didn't invalidate the amyloid hypothesis — the genetics remain real — but it introduced genuine scientific uncertainty about which specific form of amyloid accumulation matters most. It also opened up the field to criticism that the hypothesis had been protected from rigorous challenge partly because of concentrated funding and institutional commitment.

## The lecanemab and donanemab complication

And then, in 2022–2023, the picture got genuinely complicated. Lecanemab (approved in 2023) and donanemab (approved in 2024) both showed statistically significant slowing of cognitive decline in early-stage Alzheimer's patients — roughly 25–35% reduction in the rate of decline compared to placebo. That's not a cure, and the effect size is modest. But it's real.

The results can be read in two ways. Optimists say this validates the hypothesis — amyloid is causal, earlier intervention works, better drugs will work better. Skeptics note that modest benefits with serious side effects (amyloid-related imaging abnormalities — brain swelling and microbleeds — occur in 20–35% of treated patients) at enormous cost don't yet constitute a transformational therapy.

The science is genuinely unsettled. Tau pathology, neuroinflammation (the role of microglia), and vascular factors all play roles that aren't fully captured by the amyloid cascade model. Whether amyloid is cause, consequence, or coincident feature in different subtypes of Alzheimer's remains an open question. What we've learned from 30 years of trials is mostly that the disease is more heterogeneous than the original hypothesis assumed.

That's not nothing. But it's also not a cure.

The Alzheimer's Science Problem: Why the Amyloid Hypothesis Keeps Failing Drug Trials

// COMMENTS

ON THIS PAGE