Cochrane Review of Anti-Amyloid Alzheimer's Drug Trials Draws Sharp Criticism From Field
Key Takeaways
- Seventeen placebo-controlled trials across seven antibodies yielded statistically significant but clinically small mean effects on ADAS-Cog and CDR-SB, with low-to-moderate certainty and scale-dependent functional outcomes.
- Safety signals were dominated by increased ARIA-edema risk, whereas symptomatic ARIA was less frequent and 18-month serious adverse events and all-cause mortality showed no excess.
A systematic review pooling data from 17 clinical trials—including 15 involving failed or withdrawn therapies—concludes that amyloid-targeting antibodies offer little clinically meaningful benefit, prompting widespread pushback from Alzheimer's researchers.
A newly published Cochrane systematic review examining amyloid-beta-targeting monoclonal antibodies in Alzheimer's disease has concluded that the drug class offers little to no clinically meaningful benefit in people with mild cognitive impairment or mild dementia.1
The finding has drawn immediate and pointed criticism from researchers across the field, most of whom take issue not with the question the review asks, but with how it was designed to answer it.
Published April 16 in the Cochrane Database of Systematic Reviews, the analysis assessed 17 randomized controlled trials involving 20,342 participants conducted between 2014 and 2024. The trials evaluated seven different antibodies: aducanumab, bapineuzumab, crenezumab,
Pooled data showed that amyloid-targeting antibodies “probably result in little to no difference in cognitive function” as measured by the Alzheimer's Disease Assessment Scale‐Cognitive (ADAS-Cog) scale, with a standardized mean difference of -0.11 (95% CI -0.16 to -0.06; moderate certainty).
For dementia severity on the Clinical Dementia Rating Scale Sum of Boxes (CDR-SB) scale, the treatments “may result in little to no difference,” with a standardized mean difference (SMD) of -0.12 (95% CI -0.24 to -0.00; low certainty). Effects on functional ability ranged from little to no difference to small increases depending on the scale used.
On safety, the review found a probable small increase in amyloid-related imaging abnormalities involving edema, with an absolute risk difference of 107 more cases per 1,000 participants compared to placebo. Symptomatic amyloid‐related imaging abnormalities (ARIA) was less common. No increase in serious adverse events or overall mortality was found at 18 months, both rated as high certainty evidence.
The authors concluded that successful removal of amyloid from the brain does not appear to be associated with clinically meaningful effects in this population and called for future Alzheimer's research to focus on other mechanisms of action.
A methodology under fire
The core objection from outside researchers centers on the review's decision to pool results across all seven antibodies, combining trials of drugs that failed development alongside those of two treatments now approved and in clinical use.3
Of the 17 trials included, 12 evaluated drugs that did not meet their primary endpoints. Three assessed aducanumab, which failed one of its two pivotal trials and was later withdrawn from the market. Only two trials evaluated currently available treatments: one for lecanemab and one for donanemab, both now approved in the US and other countries.
According to a media roundup shared on Science Media Centre, Bart De Strooper, group leader at the UK Dementia Research Institute at University College London (UCL), was among the most direct in his critique.
"By mixing failed drugs with the only antibodies that have actually changed clinical practice, it turns therapeutic progress into statistical noise," he said. "Once failed and successful programs are combined into a single pooled estimate, the average will inevitably look weaker than the best performing agents. That is not a biological insight—it is simply the arithmetic consequence of mixing negative and positive studies together."
The conflation of therapies with fundamentally different mechanisms was another recurring objection. John Hardy, also at the UK Dementia Research Institute at UCL, pointed out that aducanumab works by removing existing plaques while lecanemab primarily binds to soluble forms of amyloid to prevent plaque formation—making them distinct treatments, not interchangeable data points.
That distinction matters for how results are interpreted, several researchers noted. Because 15 of the 17 trials included drugs no longer used in clinical practice, the combined results do not accurately reflect the two treatments now approved and in active use, according to Suzanne Schindler, MD, PhD, of Washington University in St. Louis, in comments made to MedPage Today.
Combining studies of multiple drugs, many of which had little or no effect on beta-amyloid and most of which failed in randomized trials, makes an overall finding of clinical ineffectiveness almost inevitable, Jonathan Schott, professor of neurology at UCL, argued—while not ruling out meaningful benefit from individual agents.
The small magnitude of benefit seen across individual trials was also worth putting in context, according to Robert Howard, professor of old age psychiatry at UCL. Those benefits generally amounted to around 2% of the range of changes measured by validated cognitive tests and achieved statistical significance only because of very large participant numbers, without reaching accepted levels of clinical efficacy—a limitation that predates this review.
What comes next
Researchers were careful to separate criticism of the review's methodology from broader conclusions about the field. Regulatory analyses by the EMA and the UK Medicines and Healthcare products Regulatory Agency continue to support the use of lecanemab and donanemab to slow decline in early Alzheimer's disease, Tara Spires-Jones, professor of neurodegeneration at the University of Edinburgh noted.
However, the findings do reinforce a growing view that amyloid removal alone is not sufficient to significantly improve cognition or slow disease progression. B. Paul Morgan, director of UK DRI Cardiff at Cardiff University, called for research into why the disease continues to advance after successful amyloid clearance, pointing to the need for complementary therapeutic targets.
One variable the review was unable to account for is the APOE4 genotype, which could not be examined in subgroup analysis because genetic testing is not standard clinical practice. Andrea Kwakowsky, PhD, associate professor in pharmacology and therapeutics at the University of Galway, described it as potentially significant for understanding both therapeutic response and the risk of brain swelling and bleeding associated with this drug class.
Longer-term evidence is also missing from the picture. Susan Kohlhaas, PhD, executive director of research at Alzheimer's Research UK, noted that newer data suggesting the approved treatments may deliver modest but sustained benefits beyond 18 months is not captured in the review, and called for further study through real-world healthcare settings to better understand how these treatments perform in practice.
The review's authors acknowledged that the external validity of their findings is limited by strict trial inclusion criteria, with estimates suggesting only 5% to 15% of the broader Alzheimer's population would have been eligible for the included studies, and that longer follow-up data will be needed to fully characterize the long-term safety profile of this drug class.
References
1. George J. Alzheimer's Drug Review Ignites Backlash From Experts. MedPage Today. April 16, 2024. Accessed April 22, 2024.
2. 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 of Systematic Reviews 2026, Issue 4. Art. No.: CD016297. DOI: 10.1002/14651858.CD016297. April 16, 2026. Accessed April 22, 2026.
3. expert reaction to Cochrane review of anti-amyloid monoclonal antibodies for Alzheimer’s Disease. Science Media Centre. April 16, 2026. Accessed April 22, 2026.
Related Content
