Amyloid's Reckoning: Cochrane Review Finds Alzheimer's Drugs' Benefits "Absent or Trivial" as Medicare Bills Mount

On April 16, 2026, Cochrane — the global collaborative that produces the reference-standard systematic reviews used by regulators and clinicians worldwide — published an analysis that is forcing a reckoning in Alzheimer's medicine. Pooling data from 17 randomized trials covering 20,342 participants with mild cognitive impairment or mild dementia, the review concluded that amyloid-beta-targeting monoclonal antibodies "probably result in little to no difference" on cognition or dementia severity at 18 months, with absolute effects that were "absent or trivial" and fell below accepted thresholds for a minimum clinically important difference [1][2][3].

The finding lands in the middle of a commercial rollout. Eisai and Biogen's lecanemab (Leqembi) and Eli Lilly's donanemab (Kisunla) are being infused into American patients at a list price of roughly $26,500 per year, a sum the Centers for Medicare & Medicaid Services actuaries have projected will cost the Medicare program about $3.5 billion in 2025 alone [4][5]. Lead author Francesco Nonino summarized the authors' position bluntly: "Unfortunately, the evidence suggests that these drugs make no meaningful difference to patients" [1].

Source: FDA clinical review / CLARITY-AD / TRAILBLAZER-ALZ 2

Data as of Jul 2, 2024CSV

The Threshold Argument

The dispute at the center of the review is not whether the drugs do something — they visibly clear amyloid plaques from the brain on PET imaging — but whether what they do crosses the line separating statistical significance from a benefit that a patient, caregiver or physician could actually perceive.

The outcome measure at issue is the Clinical Dementia Rating Scale–Sum of Boxes (CDR-SB), an 18-point scale on which higher scores indicate more severe dementia. Analyses of the minimum clinically important difference (MCID) typically set the threshold at about 1 point for patients with mild cognitive impairment and 2 points for mild Alzheimer's dementia [6]. In the CLARITY-AD trial that supported lecanemab's full FDA approval, patients receiving the drug declined by 1.21 points over 18 months versus 1.66 points on placebo — a 0.45-point absolute difference, roughly a 27% relative slowing [6][7]. Donanemab in TRAILBLAZER-ALZ 2 showed a 0.67-point absolute advantage, a 29% relative slowing [6][7]. Both absolute differences sit below the 1-point MCID for MCI.

The FDA nevertheless characterized both effects as "clinically meaningful" in its approval decisions, relying in large part on the consistency and statistical strength of the results across the CDR-SB's subdomains [7]. Regulators also distinguished between standards: aducanumab's 2021 accelerated approval rested on a biomarker — amyloid clearance — as a surrogate endpoint, while lecanemab's July 2023 traditional approval and donanemab's July 2024 traditional approval required demonstration of clinical benefit on cognitive/functional scales [8][9]. That pivot, from surrogate to clinical endpoint, followed intense criticism of the aducanumab decision, which drew a 10–0 vote against approval from the FDA's own Peripheral and Central Nervous System Advisory Committee and prompted the resignation of three panel members [8].

A Defense Built on Subgroups

The review's conclusions drew pointed rebuttal from researchers associated with the amyloid hypothesis and its industry sponsors. Professor Sir John Hardy of University College London — whose consulting history includes Lilly, Biogen and Eisai and who first articulated the amyloid cascade hypothesis in the early 1990s — called the Cochrane analysis "a silly paper which should not have been published," arguing that pooling the two currently approved drugs with failed predecessors like bapineuzumab and crenezumab mathematically drags the average toward null [3][10]. Professor Bart De Strooper of the UK Dementia Research Institute argued the review "turns therapeutic progress into statistical noise" by mixing failed drugs with antibodies "that have actually changed clinical practice" [10].

The steelman defense goes further than methodology. A Bayesian reanalysis of phase 3 data published in 2025 found that lecanemab and donanemab were effectively ineffective in APOE ε4 homozygotes — the highest-genetic-risk group — but retained measurable effects in non-carriers and heterozygotes [11]. That signal carried regulatory weight abroad: the European Medicines Agency ultimately recommended lecanemab only for non-carriers and heterozygotes, and the UK's Medicines and Healthcare products Regulatory Agency adopted an equivalent restriction [11]. Defenders of the drugs argue that aggregate meta-analyses obscure subgroup benefits that are genuine and clinically meaningful for specific patients — particularly those identified early, in the MCI stage, with lower baseline tau burden [10][12].

Other academic reactions split the difference. Professor Ian Maidment at Aston University agreed the drugs "may show a statistical effect" but said it "was unlikely to be clinically relevant for patients," while Professor Andrew Doig of Manchester concluded "the effect is so small that it would never be noticed in practice" [10]. Susan Kohlhaas of Alzheimer's Research UK warned against dismissing the impact as "trivial" given the analytical constraints, while the Alzheimer's Society's Richard Oakley said the review "makes the picture look bleaker than it really is" [10].

The Safety Ledger

If the benefit side of the equation is contested, the harms are not. Anti-amyloid antibodies produce a class-specific imaging finding called ARIA — amyloid-related imaging abnormalities — comprising edema (ARIA-E) and microhemorrhages or superficial siderosis (ARIA-H). Most events are asymptomatic and detected only on surveillance MRI, but symptomatic and fatal cases have occurred.

Source: Neurology meta-analysis / FDA labeling

Data as of Dec 1, 2024CSV

In phase 3 trials, ARIA-E occurred in roughly 12.6% of lecanemab-treated patients and ARIA-H in 17.3%; for donanemab the rates were approximately 24% for ARIA-E and 31.4% for ARIA-H [13][14]. Aducanumab's rate was higher still at 35.2% for ARIA-E [13]. APOE ε4 carrier status roughly doubles ARIA-E risk and triples ARIA-H risk compared with non-carriers, with homozygotes facing the highest rates [15]. The FDA's post-marketing analysis identified six fatal ARIA-E cases tied to lecanemab, with five patients developing symptoms within 0–8 days of their most recent infusion [16]. A separate adverse event database review identified 101 serious ARIA-E cases, the majority clustering between the third and fifth infusions [16]. The FDA subsequently recommended additional, earlier MRI monitoring [16]. One ARIA-H fatality has been reported with donanemab and one ARIA-E death with aducanumab [13].

The severity-benefit calculus tightens considerably if the Cochrane conclusions are correct. Weighing a ~0.45-point CDR-SB shift against a roughly one-in-five chance of brain swelling or micro-bleeding — and a small but non-zero chance of a fatal event — is different from weighing a clinically perceptible cognitive preservation against the same risks.

Money, Disclosures and the Approval Process

The commercial stakes are large and expanding. A 2023 JAMA Internal Medicine projection estimated $2 billion to $5 billion in annual Medicare spending on lecanemab and its ancillary costs — PET imaging, serial MRIs, infusion center time, APOE testing — once prescribing scaled [17]. CMS's own actuaries forecast about $3.5 billion across the Medicare program in 2025 [4]. Because lecanemab and donanemab are administered in clinical settings, they are covered under Medicare Part B, which typically pays 80% after the deductible, leaving substantial out-of-pocket exposure for beneficiaries without supplemental coverage [5].

Source: CMS Office of the Actuary / JAMA Internal Medicine / KFF

Data as of Apr 11, 2024CSV

Aducanumab's short commercial life illustrates the pattern in concentrate. Approved in June 2021 at an annual list price of $56,000, it was later cut by Biogen to $28,200, blocked from routine Medicare coverage in April 2022, and formally discontinued in January 2024 [8][18]. Biogen stated the decision was "not related to any safety or efficacy concerns" and redirected resources to lecanemab [18].

Financial conflicts among the regulators' outside advisers have drawn scrutiny. A September 2024 BMJ investigation by Jeanne Lenzer and Shannon Brownlee reported that seven of eight members of the FDA advisory committee that reviewed donanemab had received direct payments from drug manufacturers, with individual advisers receiving consulting and speaking fees of up to $62,000 and research grants of up to $10.5 million between 2017 and the end of 2023 [19]. Three advisers had financial ties to Lilly, two to Roche (Lilly's development partner for an Alzheimer's blood test) and two others held patents on amyloid antibodies, according to the BMJ report [19]. The investigation did not establish that those relationships changed any vote, but it sharpened long-running questions about how clinical endpoints for Alzheimer's trials — particularly the choice and interpretation of CDR-SB thresholds — were negotiated between sponsors and regulators.

Historical Precedent: When Disease-Modifying Bets Fail

The amyloid hypothesis has dominated Alzheimer's drug development for more than three decades, with hundreds of failed candidates preceding lecanemab and donanemab [20]. Critics draw parallels to other disease-modifying programs that were ultimately abandoned after years of investment — for instance, HDL-raising drugs for cardiovascular prevention, where torcetrapib and its successors produced biomarker effects without corresponding mortality benefit. If the Cochrane interpretation holds, amyloid monoclonal antibodies would join that list of mechanistically elegant therapies that moved a biomarker but failed to move patients.

The field has already started diversifying. National Institute on Aging data show that NIH's translational Alzheimer's portfolio has shifted such that more than 60% of funding, and about 70% of basic research funding, now targets mechanisms other than amyloid or tau [21]. The 2025 pipeline contains 138 drugs across 182 trials; only 22% of phase 3 agents target amyloid and 3% target tau, with the remainder spread across neuroinflammation, synaptic function, metabolism and other mechanisms [22].

Source: OpenAlex

Data as of Jan 1, 2026CSV

What Happens Next to Patients, Prescribers and Pipelines

Roughly 6.9 million Americans age 65 and older were living with Alzheimer's in 2024, with that figure rising to an estimated 7.2 million in 2025; unpaid caregivers provided 19 billion hours of care valued at more than $413 billion in 2024 [23]. Surveys of those caregivers find their views on anti-amyloid therapy to be far more muted than the "breakthrough" framing used in public communications: one analysis reported that 56.9% of caregivers preferred insurance coverage for home-based care over medication, and most perceived only "modest clinical value" in the new agents [24].

For prescribing physicians, the Cochrane analysis arrives alongside FDA-required black-box-style warnings about ARIA and a tightened MRI monitoring schedule. If benefit is absent for the average patient in the label population — and meaningfully present, as some defenders argue, only for genetically-stratified subgroups — informed consent discussions become more complicated, and the liability exposure for patients who experience symptomatic ARIA without cognitive gain becomes more exposed. That legal exposure is not theoretical: the FDA has documented fatalities, and product-liability precedent in the United States has historically tracked whether a manufacturer's benefit claims proved reconcilable with the real-world risk profile.

The pipeline is already repositioning. Roche's trontinemab, engineered with "brain shuttle" technology to cross the blood-brain barrier at higher concentrations with potentially lower ARIA risk, entered phase 3 TRONTIER trials in 2025 with early biomarker data expected in 2026 [22]. Biogen's BIIB080, a tau-targeting antisense oligonucleotide, completed enrollment in its phase 2 CELIA trial in 2025 with readouts expected in 2026 [22]. Combination approaches — simultaneously targeting amyloid, tau and neuroinflammation — are emerging as a consensus strategy among academic groups, including work on senolytic combinations such as dasatinib and quercetin in preclinical models [25].

The Unresolved Question

The Cochrane authors and their defenders are not disputing the same data; they are disputing which comparisons are legitimate. If the approved drugs should be evaluated against placebo only in their own trials — CLARITY-AD and TRAILBLAZER-ALZ 2 — the absolute effect on CDR-SB is small but statistically clear, and reasonable regulators have called it clinically meaningful [7]. If they should be evaluated against the MCID threshold derived from broader populations, or against the null hypothesis after pooling with failed amyloid antibodies on the grounds that all target the same molecule, the effect collapses to "trivial" [1][3].

That ambiguity is not new in drug evaluation, but the stakes here are high. A class of medicines now consumes several billion dollars a year in public and private insurance spending, exposes roughly one in five recipients to brain imaging abnormalities and forces a generation of patients and families to weigh surveillance MRIs and infusion schedules against a cognitive effect that some of the world's most authoritative evidence-synthesizers now say they cannot reliably detect. Whether regulators will revisit approvals, whether payers will tighten coverage, and whether the field will redirect more rapidly toward tau, inflammation and combination approaches are questions the Cochrane review has made unavoidable — even if it has not, on its own, resolved them.