Alzheimer’s Drugs Clear the Plaques but Leave Patients No Better Off

The brain scans look unambiguous. After 18 months of treatment, the amyloid plaques that riddle the brains of early Alzheimer’s patients are visibly reduced, sometimes dramatically so. The drug is doing exactly what it was designed to do: hunting down those sticky protein deposits and clearing them away. So why aren’t the patients getting better? That question, hovering over Alzheimer’s research for the better part of two decades, has now received its most definitive answer yet, and it’s not the one the pharmaceutical industry was hoping for.

A major Cochrane review published today, pooling data from 17 clinical trials and more than 20,000 patients, concludes that anti-amyloid monoclonal antibodies produce no clinically meaningful improvement in memory, daily functioning, or overall dementia severity. Meanwhile, they reliably cause brain swelling and microbleeds in a substantial minority of users.

The Gap Between Statistics and Reality

The anti-amyloid drugs (a class that includes lecanemab, donanemab, aducanumab, and several earlier candidates that never made it to market) were built on the amyloid cascade hypothesis, which holds that the waxy plaques of aggregated amyloid-beta protein are not just a feature of Alzheimer’s disease but its driving cause. Remove the plaques, the theory goes, and you slow the disease. It’s a logical enough premise, and for years it commanded enormous financial and scientific investment. The drugs that resulted do, genuinely, clear amyloid. That part works. What doesn’t follow is the next step in the chain of reasoning: that this clearance translates into anything a patient can actually feel.

The numbers tell a blunt story. On the standard cognitive assessment scale used across almost all 17 trials, the treatment groups scored about 0.85 points better than placebo after 18 months. The minimum difference considered clinically meaningful (the smallest change a patient or caregiver would notice) is somewhere between two and four points. A similar gap yawns open in dementia severity scores. “Unfortunately, the evidence suggests that these drugs make no meaningful difference to patients,” said lead author Francesco Nonino, a neurologist and epidemiologist at the IRCCS Institute of Neurological Sciences of Bologna.

Nonino is careful to separate the statistical from the clinical. Individual trials produced results that were technically significant, meaning they were unlikely to have occurred by chance, and those results generated considerable excitement when they were published. But statistical significance is not the same thing as clinical relevance, and in a disease as devastating as Alzheimer’s, the distinction matters enormously. “It is common for trials to find statistically significant results that do not translate into a meaningful clinical difference for patients,” Nonino noted.

Brain Swelling and the Costs of Treatment

The adverse event picture complicates things further. Amyloid-related imaging abnormalities (the clinical euphemism for brain swelling and microbleeds detected on MRI) occurred in roughly 119 out of every 1,000 patients taking the drugs, compared to about 12 per 1,000 in the placebo groups. Most of these events showed no obvious symptoms and were caught only on routine scans, though their long-term significance remains genuinely unknown; the trials simply didn’t run long enough, and reporting of symptomatic cases was inconsistently handled across studies.

Patients carrying the APOE ε4 gene variant (a known risk factor for Alzheimer’s) face an even higher rate of these brain changes, a finding that has real implications given how common this variant is in people who develop the disease. Add to this the logistical burden of the treatment: intravenous infusions every two to four weeks, baseline PET scans or spinal fluid tests to confirm amyloid positivity, and repeated MRI monitoring for adverse events. In practical terms, the therapy is accessible only to well-resourced healthcare systems, and even there, not equitably.

The Deeper Problem With the Hypothesis

What the review’s authors find most significant isn’t just the clinical null result. It’s the fact that the drugs achieve their stated biomarker target while failing to move the needle on what patients actually care about. That disconnect is hard to explain away, and several of the included trials were themselves stopped early for futility. A phase 3 trial of solanezumab in people who were amyloid-positive but not yet symptomatic (a population chosen precisely because the theory predicts earlier intervention should work better) also showed no benefit. Similarly, a long-running trial in dominantly inherited Alzheimer’s disease found no clinical gains despite successful amyloid removal after nearly five years of treatment. The converging weight of evidence, across different drugs, different patient populations, and different trial designs, suggests the amyloid hypothesis may be missing something fundamental about how the disease actually unfolds.

This matters because the hypothesis has, until recently, structured almost the entire disease-modifying drug pipeline for Alzheimer’s. Some researchers have argued for years that amyloid is more bystander than culprit, or that it acts early in the disease but has already triggered irreversible downstream damage by the time symptoms appear. The Cochrane authors note there are now quite a few other directions worth pursuing: neuroinflammatory pathways, the relationship between the gut microbiome and brain function, vascular risk factors, and tau protein tangles, which represent a distinct pathological hallmark of the disease. None of these has yet produced an approved treatment, but none has been tested at anything like the same scale as amyloid-targeting.

For Edo Richard, Professor of Neurology at Radboud University Medical Centre and senior author on the review, the findings carry a personal weight. “I see Alzheimer’s patients in my clinic every week and I wish I had an effective treatment to offer them,” he said. The drugs that currently exist for Alzheimer’s, older agents that modulate neurotransmitter systems, provide modest symptomatic benefit for some patients but do nothing to slow the disease’s progression. Anti-amyloid antibodies were supposed to be different in kind, not just degree. “Sadly, anti-amyloid drugs do not offer this and bring additional risks,” Richard said. “Given the absence of correlation between amyloid removal and clinical benefit, we need to explore other pathways to help address this devastating disease.”

There’s also a risk, the authors acknowledge, that some of the modest apparent benefits seen in individual trials may be inflated. Because the drugs cause distinctive side effects (infusion reactions, the brain changes visible on scans), both patients and their carers may have been able to guess which arm of a trial they were in, potentially biasing their self-reports of functioning and cognition. It’s not a certainty, and it doesn’t change the overall picture much, but it adds another layer of uncertainty to numbers that were already well below what anyone would call clinically meaningful.

Regulatory bodies have reached strikingly different conclusions about all this. The FDA and the European Medicines Agency have approved lecanemab and donanemab; the UK regulator followed suit. But England’s National Institute for Health and Care Excellence refused to recommend NHS reimbursement for both drugs, concluding the benefits were too small to justify the cost. The Dutch and German health assessment bodies reached similar conclusions. These divergences reflect genuine disagreement about what constitutes acceptable benefit in a disease with no good alternatives, and that disagreement is now considerably harder to sustain in the face of a 20,000-patient Cochrane synthesis.

Six additional trials are ongoing. Some are testing the same drugs at different stages of disease; others are exploring combination approaches. It remains just possible that an earlier intervention window, or a patient population selected more precisely by genetic or biomarker profile, might yet reveal a meaningful clinical signal. But the Cochrane authors are measured about that prospect. The biology of amyloid removal and the biology of cognitive preservation, the data now suggest, may be far more loosely coupled than the hypothesis predicted, and the disease that has eluded treatment for so long may require researchers to look somewhere else entirely.

---

Source: Cochrane Database of Systematic Reviews. DOI: 10.1002/14651858.CD016297

Frequently Asked Questions

If the drugs actually clear amyloid from the brain, why don’t patients improve?

That’s the central puzzle the review highlights. Anti-amyloid antibodies do what they’re designed to do: they measurably reduce amyloid plaque burden. But this biological change doesn’t appear to translate into slower cognitive decline or any improvement patients can feel. The disconnect suggests the amyloid hypothesis may be incomplete: amyloid might be a marker of disease rather than its engine, or its removal may come too late in a cascade of damage already set in motion years earlier.

Are the brain side effects from these drugs serious?

They’re real but complex to assess. Brain swelling and microbleeds (collectively called ARIA) show up on MRI in roughly one in ten treated patients, compared to about one in a hundred on placebo. Most cases appear to produce no obvious symptoms. The concern is that long-term consequences of these changes are genuinely unknown, the trials weren’t long enough to find out, and reporting of symptomatic cases was inconsistent across studies, so patients and doctors are making decisions with incomplete safety information.

Does this mean the amyloid hypothesis for Alzheimer’s is wrong?

Not necessarily wrong, but the evidence is increasingly awkward for its strongest form. The hypothesis holds that amyloid accumulation drives the disease, so clearing it should slow progression. Seven drugs have now been tested in large trials; none has produced a clinically meaningful benefit. Researchers are increasingly interested in other mechanisms: neuroinflammation, tau protein pathology, gut-brain interactions, though none has yet been tested at comparable scale.

Why have some regulators approved these drugs if the evidence is this weak?

Regulatory approval and clinical benefit aren’t always the same threshold. The FDA and European Medicines Agency approved lecanemab and donanemab partly on the basis that the disease has no effective alternatives and the drugs do produce measurable biomarker changes. Health assessment bodies focused on cost-effectiveness; England’s NICE and equivalent agencies in the Netherlands and Germany rejected reimbursement, concluding the benefit too small to justify the expense and monitoring burden. The Cochrane review, with its larger pooled dataset, strengthens the case for the sceptical position.