Eight days ago, on Thursday, April 16, the Cochrane Collaboration published a systematic review of seventeen clinical trials of anti-amyloid monoclonal antibodies for Alzheimer's disease. The review pooled data from 20,342 participants and concluded that the absolute effects on cognitive decline and dementia severity were "absent or trivial," falling well below the established threshold for a minimum clinically important difference, while the drugs increased the risk of brain swelling and bleeding. [1] Lead author Francesco Nonino of Bologna's IRCCS Institute of Neurological Sciences said the evidence had converged: early statistically-significant trial results did not translate into clinical benefit. [2] The paper's Tuesday Day Seven account called the silence from the FDA, CMS, Biogen, and Eli Lilly structural. Today is Day Eight. The silence from those four institutions has not broken, and the critique the Cochrane review triggered among working neurologists has sharpened into an open split.

The split runs along a single axis: whether the Cochrane review, by pooling seventeen trials of seven different drugs, averaged the two newest agents — lecanemab (Leqembi, Biogen/Eisai) and donanemab (Kisunla, Eli Lilly) — down into a series of older failures. The Alzheimer's Society, which in the United Kingdom has endorsed the two drugs despite their NICE rejection on cost grounds, issued a response the day after Cochrane's publication arguing that the review "combined results for a majority of failed drug trials with a small number of more recent successful trials." [3] That framing — "trial-mixing artifact" in the paper's shorthand — is the defensive argument. The opposing argument, which the Science Media Centre and the Pharmaceutical Journal have collected in the form of expert reactions, is that if six of seven drugs fail and the seventh shows statistical but not clinical significance, the class signal is the failure, not the outlier. [4][5]

Consider the arithmetic the Cochrane authors themselves put on the record. The absolute effect of anti-amyloid drugs on cognitive decline at 18 months — the trial endpoint used across all seventeen studies — sits "well below" the minimum clinically important difference. [1] The minimum clinically important difference, in this context, is the smallest change a patient or caregiver can reliably notice. A drug that clears amyloid from the brain but does not produce a noticeable clinical change is, by the standard definition, not a disease-modifying therapy. The Alzheimer's Society's rebuttal does not contest this arithmetic for the older drugs. It contests whether pooling dilutes a real lecanemab/donanemab signal.

The question of how to read the Cochrane finding is structural rather than scientific. Cochrane reviews are, by design, the slowest and most conservative form of evidence aggregation in medicine. They are the tool used to close debates, not to open them. When the finding is "absent or trivial" and the policy response is silence, the review functions as the closing argument of a trial that the regulatory and reimbursement architecture is declining to hear. AARP quoted dissenting neurologists saying eighteen months is too short to capture the benefit of a slowly-acting drug in a slowly-progressing disease; Cochrane's authors acknowledged this limitation and noted that eighteen months is the length of the randomized controlled trials that generated the entire evidence base. [6] The trials that would answer the eighteen-month question are not running.

The regulatory record continues to move in the direction opposite the evidence. On April 15, the day before Cochrane published, the FDA approved a new maintenance dosing schedule for lecanemab, allowing patients to switch to less-frequent infusions after eighteen months of standard treatment. [7] The maintenance-dosing approval was based on open-label extension data, not a randomized clinical trial, and suggests that the FDA's regulatory stance is to operationalize continued use rather than re-examine the initial approval. CMS continues to reimburse both drugs under the National Coverage Determination tied to traditional FDA approval. Medicare's Alzheimer-coverage page has not been updated to acknowledge the Cochrane review. The National Institute for Health and Care Excellence in the UK rejected both drugs on cost-effectiveness grounds before Cochrane; the pre-existing transatlantic divide — FDA-approved, NICE-rejected — is now the structural frame inside which the Cochrane review lands.

What Day Eight adds is depth on the split itself. The neurologists arguing "trial-mixing artifact" are disproportionately those who ran or consulted on the lecanemab and donanemab trials, or who practice at academic centres with amyloid-focused infusion programs. The neurologists arguing "class failure" are disproportionately health-services researchers, general neurologists, and the authors of the Cochrane review itself. The split does not map cleanly onto a scientific disagreement; it maps onto which population of clinicians bears the cost of the drug and which bears the cost of the ARIA (amyloid-related imaging abnormalities) complications. Infusion-program physicians see the marginal benefit. Emergency-medicine physicians see the brain bleeds. Both groups are looking at the same evidence base.

The FDA, CMS, Biogen, and Lilly have not spoken on Day Eight. Biogen's investor-relations page carries no update. Lilly's press releases since April 16 address unrelated products. CMS's Leqembi coverage page is unchanged. The paper continues to read the institutional silence as policy rather than delay; the duration is now past the point at which any recent peer-review-without-regulator-response could be called a response cycle. The paper's Day Seven position was that silence at this duration becomes the institutional answer. Day Eight confirms it.

The next inflection is procedural rather than scientific. If CMS initiates a National Coverage Determination reconsideration — the mechanism by which Medicare formally reviews existing coverage in light of new evidence — the thread moves to a defined timetable. If the FDA issues a label update referring to the Cochrane findings, the thread moves to a regulatory-action frame. Neither is currently on any published agenda. Until one appears, the infusion pipeline continues to operate at its pre-Cochrane rate, the amyloid PET scans continue to be ordered, and the ARIA-monitoring MRIs continue to be billed. The reimbursement decision, as the paper has now said three times, was never about the evidence. The evidence arrived last Thursday. The decision did not change.

-- NORA WHITFIELD, Chicago