A gardener who rips out every plant in the bed - roses and crabgrass alike - then concludes "nothing here was worth growing" hasn't really proven anything about roses. That, in a nutshell, is what three prominent Alzheimer's researchers just argued happened to the latest big-ticket review of anti-amyloid drugs for Alzheimer's disease. And honestly? The metaphor writes itself, because the word "aggregation" is doing double duty in ways that would make my thesis advisor proud.
The Setup: A Review That Stirred the Pot
In April 2026, a Cochrane review landed like a grenade in the Alzheimer's research world. The team, led by Francesco Nonino, analyzed 17 randomized trials of amyloid-targeting antibodies involving over 20,000 participants and concluded that these drugs "probably result in little to no difference" in cognitive decline (Nonino et al., 2026; PMCID: PMC13082890). Headlines were brutal. "New Alzheimer's Drugs Provide No Meaningful Benefit" ran one. Cue the sound of a thousand grant applications quietly weeping.
But here's where it gets interesting (I promise, and I never use that word lightly except when I absolutely do). The review pooled together seven different antibodies: aducanumab, bapineuzumab, crenezumab, donanemab, gantenerumab, lecanemab, and solanezumab. Five of those drugs failed their clinical trials. They were abandoned. Done. Finished. They are the crabgrass in our garden metaphor, and nobody is arguing they should be replanted.
The Punchline Is in the Title
Nick Fox, Susan Kohlhaas, and Jonathan Schott - heavy hitters in dementia research - published a commentary in The Lancet with possibly the best academic title of 2026: "Alzheimer's disease immunotherapy and the amyloid hypothesis: when aggregation obscures interpretation" (Fox et al., 2026; PMID: 42035783).
Read that title again. "Aggregation" refers both to the clumping of amyloid-beta proteins in the brain (you know, the whole premise of the amyloid hypothesis) and to the statistical clumping of wildly different drugs into one analysis. I did a little fist pump when I caught that. My lab mates were concerned.
Their argument is straightforward: you cannot lump 12 trials of drugs that failed with 2 trials of drugs that actually worked - lecanemab and donanemab - and then announce the whole class is a bust. That's not a meta-analysis; that's a smoothie made from one strawberry and five rocks. Of course it's going to taste bad.
What the Successful Drugs Actually Did
Lecanemab, tested in the CLARITY-AD trial, slowed cognitive decline by 27% over 18 months compared to placebo, as measured by CDR-SB scores (van Dyck et al., 2023; PMID: 36449413). Extended follow-up data out to four years showed the benefit growing over time - roughly a one-year delay in disease progression. Donanemab showed a similar ~29% reduction in its TRAILBLAZER-ALZ 2 trial.
Are these effects modest? Yes, and I will hedge appropriately here because that's what we do. These are not cures. They come with real risks, including amyloid-related imaging abnormalities (ARIA) - brain swelling and microbleeds that sound exactly as unpleasant as they are, particularly for people carrying APOE-e4 variants. The EMA actually rejected donanemab in 2025 over these safety concerns, even as the FDA approved it.
But "modest" is not "absent or trivial," which is how the Cochrane review characterized the pooled results. As Tara Spires-Jones from the University of Edinburgh pointed out, the review weakened its own conclusions by combining data from "5 drugs that did not succeed" with "2 drugs that did succeed in slowing disease progression."
Why This Matters Beyond Academic Squabbling
This isn't just scientists arguing about statistics at a conference (though we do love that). The amyloid hypothesis - the idea that toxic amyloid-beta protein buildup drives Alzheimer's - has been the dominant theory for over three decades, surviving repeated drug failures that nearly killed it. When lecanemab and donanemab finally showed clinical benefit, it felt like partial vindication: proof that hitting amyloid hard enough, early enough, and with the right molecular tool can actually slow this disease (Lancet Neurology, 2025; PMID: 40120602).
A review that conflates those breakthroughs with earlier failures risks real harm - discouraging patients from accessing available treatments, undermining funding for combination therapies, and muddying what the science actually shows. These drugs aren't silver bullets. Alzheimer's almost certainly involves tau tangles, neuroinflammation, and mechanisms we haven't fully mapped yet. But the answer to "these two drugs work modestly" is not "let's average them with five drugs that don't work and call it a day."
The Limitation I Can't Not Mention
(You knew this was coming.) Even the drugs that work have meaningful limitations. The clinical benefits, while statistically significant, sit at the edge of what individual patients and families can perceive. The risk of ARIA is real. Access is uneven, expensive, and requires regular infusions and MRI monitoring. We need better drugs, combination approaches, and probably a more complete model of what drives neurodegeneration. But we also need honest data analysis - and that means not burying your roses under a pile of weeds.
References:
-
Fox NC, Kohlhaas S, Schott JM. Alzheimer's disease immunotherapy and the amyloid hypothesis: when aggregation obscures interpretation. Lancet. 2026. DOI: 10.1016/S0140-6736(26)00789-0. PMID: 42035783.
-
Nonino F, et al. Amyloid-beta-targeting monoclonal antibodies for people with mild cognitive impairment or mild dementia due to Alzheimer's disease. Cochrane Database Syst Rev. 2026;4:CD016297. DOI: 10.1002/14651858.CD016297. PMCID: PMC13082890.
-
van Dyck CH, Swanson CJ, Aisen P, et al. Lecanemab in early Alzheimer's disease. N Engl J Med. 2023;388(1):9-21. DOI: 10.1056/NEJMoa2212948. PMID: 36449413.
-
Amyloid immunotherapy to prevent Alzheimer's disease: the wrong drug at the right time? Lancet Neurol. 2025. DOI: 10.1016/S1474-4422(25)00066-3. PMID: 40120602.
Disclaimer: The image accompanying this article is for illustrative purposes only and does not depict actual experimental results, data, or biological mechanisms.