A Cochrane systematic review covering seventeen randomized controlled trials and 20,342 participants concluded last week that anti-amyloid monoclonal antibodies for Alzheimer's disease produce no clinically meaningful benefit on memory or dementia severity, and raise the risk of brain swelling and bleeding. [1] Health Canada approved Eli Lilly's Kisunla — donanemab — on Monday, May 4, for adults with early symptomatic Alzheimer's disease. [2] The two documents are roughly one week apart. The Cochrane review is the most rigorous evidence synthesis available on the class. The Health Canada approval is the second G7 regulator to clear a member of that class for clinical use after the U.S. Food and Drug Administration's prior clearance of donanemab and lecanemab. The two arrived at opposite verdicts on the same drugs in the same news cycle.

A Tuesday-morning American reader looking at the Canadian press wakes up to a headline saying a new Alzheimer's treatment has arrived. A reader of CIDRAP or ScienceDaily wakes up to a headline saying the class does not work and may cause harm. The story is the gap between those headlines, not either headline alone.

The Cochrane review's specific findings. Across the seventeen trials examined — most of them tested anti-amyloid antibodies in patients with mild cognitive impairment or early Alzheimer's — the pooled effect size on standardized cognitive measures was below thresholds clinicians treat as clinically meaningful. [3] Patients receiving the active drug experienced statistically significant differences from placebo on amyloid plaque clearance, which is what the drugs are designed to do, but those amyloid reductions did not translate into the cognitive or functional benefits patients and families care about. The harm side of the review's ledger documented increased rates of amyloid-related imaging abnormalities — ARIA-E for edema and ARIA-H for microhemorrhage — at rates that the review's authors said warrant clinical caution. [4] The Cochrane authors' summary statement was not equivocal. "Anti-amyloid Alzheimer's drugs show no clinically meaningful effect," reads the title of the press release the Cochrane network issued.

The pushback came quickly. The UK Dementia Research Institute issued a statement that the meta-analysis "conflates failed drugs with lecanemab and donanemab" — meaning that the seventeen-trial pool included earlier compounds that had been abandoned for lack of efficacy or safety, and that pooling those failures with the two compounds the FDA and now Health Canada have approved obscures the cleaner signal those two drugs produce when analyzed alone. [5] The pushback is methodologically real. Meta-analyses of a heterogeneous class can produce a null pooled effect even when some members of the class have a real, if modest, effect. Whether the pooling decision is appropriate depends on how alike the seventeen drugs are mechanistically, and the Cochrane authors maintain that the pooling is justified because the drugs share the amyloid-clearance mechanism. The UK Dementia Research Institute disagrees. Both positions are defensible.

What is not contested. The amyloid hypothesis — that aggregated beta-amyloid plaques are the cause of Alzheimer's disease and that clearing them slows or reverses disease — has been the central organizing theory of Alzheimer's drug development for thirty years. The hypothesis has produced a string of failed compounds: solanezumab, bapineuzumab, gantenerumab, crenezumab, and others. [6] Lecanemab (Leqembi, Eisai/Biogen) and donanemab (Kisunla, Lilly) are the first two compounds in the class that produced cognitive endpoints statistically separable from placebo in pivotal trials. Both clear amyloid plaque robustly. Both produce ARIA at meaningful rates — roughly twenty percent for ARIA-E in the donanemab pivotal, with about three percent symptomatic — and both have produced patient deaths attributed to brain hemorrhage during dosing. [7] The FDA approved both. The European Medicines Agency initially rejected lecanemab citing the safety profile, then approved a restricted indication. The U.K.'s NICE has run a multi-year consultation on whether to fund either drug through the National Health Service; that consultation is still in progress and Cochrane's review now sits inside it as the freshest evidence on the table. [8]

The Health Canada approval announced Monday is for Kisunla in adults with early symptomatic Alzheimer's. [9] The label echoes the FDA label: monthly intravenous infusions; a defined treatment course tied to amyloid clearance; required MRI surveillance for ARIA. The Health Canada decision will route through provincial drug formularies, which decide whether the public payer will reimburse. The list price in the U.S. is approximately $32,000 per year before infusion costs and MRI surveillance; the Canadian pricing has not been finalized. [10] The Common Drug Review — the pan-Canadian negotiator that recommends to provincial payers — has not yet issued its recommendation on Kisunla. The Cochrane review will be one of the documents the Review's committee considers. So will the UK Dementia Research Institute's pushback. So will the FDA's accelerated-approval data and the donanemab pivotal trial results.

The clinical reading. A neurologist treating a sixty-eight-year-old patient with mild cognitive impairment and a positive amyloid PET scan reads the Cochrane review and the Health Canada approval and faces the same patient choice she faced last Friday. Donanemab is now an approved option in Canada. Cochrane says the class does not produce clinically meaningful benefit. Both can be true at once if the meaningful benefit threshold is set higher than what donanemab can clear and lower than what an individual patient may experience as worthwhile. The Clinical Dementia Rating-Sum of Boxes endpoint donanemab cleared in TRAILBLAZER-ALZ 2 was a 0.7-point difference from placebo at eighteen months on a scale where 1 point typically marks a perceptible decline. [11] Patients and families differ on whether 0.7 points over eighteen months is meaningful. So do clinicians. So do regulators. The Cochrane review's pooled effect across all seventeen trials and 20,342 participants was smaller than that.

What the press freedom of regulators looks like in this case. Health Canada read the same evidence base Cochrane did, including the donanemab and lecanemab pivotal trials, and made a decision. The decision was approval. Cochrane read the same evidence base Health Canada did, including the failed predecessors and the two newer approvals, and made a recommendation. The recommendation was caution. The two institutions are not reading different facts. They are weighing the same facts under different mandates. Health Canada's mandate is to determine whether a drug's benefit-risk profile clears the regulatory bar; Cochrane's mandate is to determine whether the published evidence supports clinically meaningful benefit. Those are not the same question.

This paper's position. The split is not anomalous. It is the structure of evidence-based medicine arriving at a hard case. The wrong reading is to take Cochrane as the final word and conclude that Health Canada erred, or to take Health Canada as the final word and conclude that Cochrane is being purist. Both are wrong. The right reading is that anti-amyloid antibodies, including the two now approved, produce small effects on average; produce meaningful brain-bleeding risk on average; and produce, in some individual patients, what those patients and their families experience as a real benefit that did not exist a decade ago. The class is not a wonder drug. It is a class of agents with marginal average benefit and real average harm, available to patients who want to try it, and now reimbursed in two countries on the basis of those tradeoffs. The institutional contest will continue. The NICE consultation will conclude. The Common Drug Review will report. The next-generation compounds — the brain-shuttle versions designed to reduce ARIA, the small-molecule competitors aimed at the same mechanism — are in clinical development and will produce their own pivotal data over the next several years.

The reader's takeaway. If your family is making a decision today about Kisunla or Leqembi, the Cochrane review is on the table and so is the Health Canada label. The conversation with the neurologist is the right venue. The drugs are not a cure. The evidence on benefit is contested in ways that matter; the evidence on risk is not contested. A 0.7-point CDR-SB difference at eighteen months is what the strongest member of the class can deliver on average. A roughly twenty percent ARIA-E rate is what the strongest member can produce on harm. Both numbers are real. Both numbers are small. The question of whether they are meaningful is, at the patient level, irreducibly individual. At the system level, the two regulators have given different answers in the same news cycle, and a reader is entitled to know that.

-- NORA WHITFIELD, Chicago