On 16 April 2026 Cochrane published a systematic review of amyloid-beta-targeting monoclonal antibodies in people with mild cognitive impairment or mild dementia due to Alzheimer's disease (Nonino and colleagues, CD016297). It pooled 17 industry-funded phase 3 randomised trials with 20,342 participants and reported results at 18 months: on cognition measured by ADAS-Cog, a standardised mean difference of -0.11 (95% confidence interval -0.16 to -0.06; 13 studies, 9,895 participants; moderate certainty), described as probably little to no difference; on dementia severity measured by CDR-SB, -0.12 (95% CI -0.24 to -0.00; 9 studies, 8,053 participants; low certainty). Amyloid-related imaging abnormalities with oedema were more common, about 107 more per 1,000 people than placebo. The lead author, Francesco Nonino, said in Cochrane's own notice that "the evidence suggests that these drugs make no meaningful difference to patients", and the authors urged the field to look beyond amyloid removal.

The objection arrived within a week and has not stopped. The review's own tally shows what the critics seize on: of the 17 trials, three tested aducanumab, four bapineuzumab, two crenezumab, four gantenerumab and two solanezumab, all of them discontinued or unlicensed, while donanemab and lecanemab, the two antibodies actually on the market, contributed one trial each. Correspondence in The Lancet on 24 April 2026, titled "when aggregation obscures interpretation", argued the review's headline conclusion is "fundamentally undermined by pooling agents" that differ in what they bind and how much plaque they remove. Individual members of the European Alzheimer's Disease Consortium made the same case in the Journal of Prevention of Alzheimer's Disease (Frederiksen and colleagues, published online 18 May 2026), and a 37-author perspective led by Heather Snyder of the Alzheimer's Association, in the July issue of Alzheimer's & Dementia, asked Cochrane's editorial board to consider an editorial response, a correction of press materials or a commissioned reply. Medscape reported on 9 July 2026 that some researchers had asked for withdrawal and that Cochrane declined.

The case for lumping is less quotable but not empty. Cochrane set out to test a class hypothesis, that clearing amyloid produces clinical benefit, and it answered that question with every eligible phase 3 trial rather than the two that worked best. Its authors report absolute effects falling below accepted thresholds for a minimum clinically important difference, a judgement that bites on the licensed drugs too: Tara Spires-Jones of the University of Edinburgh, herself a critic of the pooling, has said the approved antibodies show modest benefits that fall short of what clinicians consider clinically meaningful. Against that, she and others note that the European Medicines Agency and the UK regulator have assessed lecanemab and donanemab individually and continue to support their use in early Alzheimer's disease. Both things can be true: an average across a class can be near zero while a single agent still has a real, small effect.

What a family should take from this is narrower than either side's press coverage suggests. The review is about groups of people over 18 months, not about one person's next infusion; it did not test the licensed drugs against each other, could not stratify by APOE4 status, and does not change any licence. Whether treatment is worth its risks for a particular person, and what the ARIA figures mean for someone with a particular scan and genotype, is a conversation for that person's clinician, with the review and its critics both on the table.