Anti-Amyloid Drugs for Alzheimer's

Healthcare · §600

For the first time in the history of Alzheimer's disease, there are drugs that slow it down. Lecanemab and donanemab are antibody infusions that strip amyloid plaque from the brain and, in early-stage patients, shave roughly five months off the next eighteen months of cognitive decline. That is a real, FDA-approved effect — and one that European regulators have looked at and called too small for the risk. The risk is mostly brain swelling and small bleeds, picked up on MRI. The cost is around $26,000 to $32,000 a year for the drug alone, on top of scans and infusion-centre time. Whether to start is a real choice, not an obvious one.

Decide · Course Evidence Moderate Chapter Healthcare

The win is modest but real: about five months of preserved function over the year and a half of treatment, on a scale that measures memory, judgement, and daily independence. The cost is steep — drug list price north of $26,000 a year, monthly or biweekly infusions, brain MRIs every couple of months to check for swelling. About one patient in five (lecanemab) to one in three (donanemab) gets some degree of swelling or microbleeding on those MRIs; most never feel it, a few have been hospitalised, a small number have died. This is a clinician-led decision in a narrow window of disease, not a supplement you can try.

Alzheimer's brains accumulate sticky clumps of a protein called amyloid-beta. The clumps start as small soluble forms, grow into stringy protofibrils, and eventually pile into the hard plaques seen at autopsy. Lecanemab and donanemab are lab-built antibodies designed to grab one of those forms — lecanemab binds the soluble protofibrils, donanemab binds an aged form found inside established plaques — and flag them for the brain's clean-up cells to remove.

What's striking is how well the cleanup works. On the brain scan that measures plaque, a year of donanemab takes the average patient from heavy plaque load down by 84%; about half of patients are essentially plaque-free by twelve months and can stop the drug Sims 2023. Lecanemab does almost as much over eighteen months van Dyck 2023. The biology, in other words, does exactly what the drug was designed to do.

The gap between that biological success and the clinical effect — only about a quarter slower decline — is the most important thing to understand about this class. Amyloid is part of the Alzheimer's story, but it isn't the whole story. By the time the plaques are visible on a scan, a second protein called tau has already started killing neurons, blood vessels in the brain have started to fail, and inflammation is established. Clearing the plaques pulls one lever in a system with several. The lever is real, but it's smaller than the marketing.

What the trials actually showed

Two large trials anchor the approval. Both enrolled around 1,700 to 1,800 patients in the early stage of Alzheimer's — mild cognitive impairment or mild dementia, with brain scans or spinal fluid confirming amyloid was present. Patients were randomly assigned to drug or placebo and followed for eighteen months. The scale used to track decline is the CDR-SB — eighteen years of clinical use behind it, scored zero to eighteen, where every point is roughly the loss of a real-life function: remembering an appointment, managing money, getting home from the shop.

Translate that into the patient's life. A 0.45-point shaving on the CDR-SB over eighteen months is roughly five months of preserved function. The treated 70-year-old reaches the disability state at month 22 or 23 that the placebo group reached at month 18 — the appointment they would have started forgetting in spring, they start forgetting in early autumn instead. The trajectory bends; it does not reverse.

This is where reasonable people split. The proponents argue the curves keep separating in the open-label extension at three years, so eighteen months understates the eventual delay, and that five months at the start of decline buys time that compounds into preserved independence later Petersen 2023. The skeptics point out that the 0.45-point difference is half the most-cited threshold for a change a patient would actually notice, that a class-level meta-analysis pooling all the anti-amyloid drugs finds an average effect of only 0.24 points, and that the European regulator looked at the same numbers and decided the benefit didn't outweigh the harm of brain bleeds Alves 2023 EMA 2025. The US regulator approved both drugs anyway, on traditional approval, after confirmatory data FDA 2023 FDA 2024. Both reads are honest reads of the same data.

Who actually qualifies

The eligible population is much narrower than "people with Alzheimer's." To start either drug, you need all of the following: a diagnosis of mild cognitive impairment or mild dementia caused by Alzheimer's; a brain scan (amyloid PET) or spinal fluid test confirming amyloid is actually there; a clean baseline brain MRI without too many old microbleeds; and the absence of blood thinners that would make brain bleeds catastrophic. The diagnostic workup typically takes six to twelve weeks before the first dose.

Of the roughly 6.9 million Americans living with Alzheimer's, only the 10–15% in the early symptomatic window are even candidates. A meaningful slice of those then fail biomarker confirmation — somewhere between one in ten and one in three patients diagnosed clinically with Alzheimer's turn out not to have amyloid in the brain, and an amyloid-targeting drug cannot help them. A second slice fails the MRI gate. A third slice is on warfarin or a DOAC for atrial fibrillation and cannot safely combine the two. The realistic eligible population in the US sits somewhere around 1 to 1.5 million people.

The APOE genotype question

Roughly one Alzheimer's patient in seven carries two copies of a gene variant called APOE-ε4. Those patients respond about as well to the drug, but the rate of brain swelling more than triples for them — roughly one in three on lecanemab gets some degree of swelling on a monitoring MRI, and on standard-dosed donanemab the rate climbs above 40% Hampel 2023. The European regulator excluded them entirely from approved use of lecanemab; the US regulator allows it with informed consent and recommends genetic testing first so the patient and family know what they're signing up for Cummings 2023. For donanemab, a slower dose ramp tested in the TRAILBLAZER-ALZ 6 trial cut the swelling rate in these patients from 57% back down to 19% — a meaningful change in the risk math Wang 2025.

What treatment actually looks like

Both drugs are intravenous infusions given in a hospital outpatient or specialised infusion centre. Schedules differ:

Brain MRIs are mandatory throughout. Lecanemab requires scans before the 5th, 7th, and 14th infusions, plus more if the patient carries APOE-ε4 Cummings 2023. Donanemab requires scans before the 2nd, 3rd, 4th, and 7th. The scans look for the two patterns of ARIA: ARIA-E (brain swelling, usually in the back of the head, usually silent) and ARIA-H (small bleeds and iron deposits along the brain surface).

Donanemab is the first drug in the class with a planned stop date tied to a biomarker. For patients who tolerate it, the regimen looks a lot more like a course of chemotherapy than a lifelong medication — a defined start, a measurable goal, and a finish line.

When the drug is the wrong move

The headline side effect is ARIA — brain swelling and microbleeds. In the lecanemab trial, ARIA-E (swelling) appeared in 12.6% of treated patients and ARIA-H (bleeds) in 17.3%; most were silent and resolved on pausing the drug. About 2.8% had swelling severe enough to cause symptoms — usually headache, confusion, or visual changes van Dyck 2023. On donanemab the rates were higher: 24% swelling, 31% bleeds, with serious ARIA in 1.6% of patients and three deaths attributable to ARIA in the treatment arm Sims 2023.

The drugs are also not approved and not studied in moderate or severe Alzheimer's, in dementia from other causes (vascular, frontotemporal, Lewy body), or in people without amyloid on confirmatory testing. Outside the approved window, the risks remain and the benefit is unproven.

What the headlines get wrong

"It reverses Alzheimer's." It doesn't. The trajectory bends — decline runs about a quarter slower. Patients still decline.
"It works for any kind of dementia." It doesn't. It is approved only for confirmed early Alzheimer's disease. Vascular dementia, frontotemporal dementia, Lewy body dementia — all unaffected, and treatment in those patients carries the same risks with no expected benefit.
"Once the plaque is gone, the patient is cured." Donanemab clears about 84% of plaque, and the resulting clinical effect is still a 22–35% slowing — not a stop, not a reversal. Plaque is one factor. Tau pathology and brain inflammation are running on their own clocks.
"The brain swelling is just a number on a scan." Mostly it is — most ARIA is silent and resolves. But a small percentage of patients are hospitalised, a smaller percentage have lasting consequences, and a small number have died. The risk is real, not statistical.

Cost, access, and the year of your life it takes

List price for lecanemab is $26,500 a year; donanemab is around $32,000 for the full course, which usually runs about a year. Those are the drug numbers. Add baseline amyloid PET (commonly $5,000–$7,000 if not covered), APOE genotyping, the diagnostic neurology workup, and four to seven monitoring MRIs in the first year, and the system cost per patient in year one runs past $40,000.

In the US, Medicare Part B covers both drugs because they are given as infusions, but only when the prescribing clinic enrols the patient in a tracking registry that reports outcomes back to CMS CMS 2023. Even with coverage, a typical Medicare patient pays around $5,300 a year in coinsurance before any supplemental policy kicks in. Medicare alone is projected to spend $3.5 billion on lecanemab in 2025.

Outside the US the picture is different. The UK's NICE rejected lecanemab as not cost-effective for the NHS, so it's available privately only. The European regulator approved lecanemab with the APOE-ε4 homozygote exclusion and rejected donanemab outright on benefit-risk grounds EMA 2025. Canada and Australia track the US more closely.

The time burden is heavier than the price tag suggests. On lecanemab a patient and caregiver are at an infusion centre 26 times a year, plus several MRIs, plus follow-up appointments. Most of those visits anchor a half-day with travel and observation time. Donanemab is lighter — monthly for about a year, with a real finish line. For families already managing early-dementia daily life, the cumulative load is the question that often decides the call.

What the win actually looks like

The honest version: over the eighteen months of a course, the average treated patient stays themselves for about five months longer. That phrase is doing a lot of work. The five months are the difference between the trajectory and a slightly shallower trajectory — the spouse who used to come home from groceries with three of the four items still comes home with three out of four, instead of two, for a season longer. The grandkids' names take a moment longer to dissolve. The morning routine works on its own for a few more weeks. The driver's licence question gets pushed from one family meeting to the next.

On the higher end — early disease, lower tau burden, donanemab — the slowing climbed to 35% and the risk of moving to the next stage of dementia fell by 39% in the trial Sims 2023. For a family already braced for moderate dementia inside two years, pushing that boundary a season or two further is not nothing. Whether it is the right trade against a year of infusion appointments, MRI anxiety, and a real shot at a brain bleed is the decision the patient and family are actually making.

What the trial cannot tell you is whether the lines keep separating after eighteen months. The extension data look like they do; the controls are gone, so the certainty is gone too Petersen 2023. The honest framing is that a treated patient buys a known small benefit over the trial window, and a hoped-for compounding benefit beyond it, against a known concrete risk during treatment.

Adjacent territory worth knowing about: amyloid PET and CSF biomarker testing, the diagnostic gate this whole class of drug sits behind; APOE genotyping, both for the risk-stratification step in this decision and for the broader long-term risk picture; cholinesterase inhibitors and memantine, the older symptom-managing drugs that remain the standard symptomatic care alongside or instead of anti-amyloid therapy; multidomain lifestyle prevention (the FINGER-style exercise, diet, cognitive-training programmes shown to slow decline in at-risk groups); and caregiver support and advanced directives, the work that runs in parallel to any disease-modifying decision in dementia.

Related in the handbook

— Your APOE status shapes both eligibility and the brain-bleed risk of these drugs — test before deciding.
— Before reaching for an expensive, risky drug, clear out the everyday meds quietly fogging the same brain.
— Slows decline sounds big until you read the actual size — about five months over eighteen.
— Cheaper than any drug here: the shingles vaccine appears to lower dementia risk by roughly 20%. Worth it from 50 on.
— Among the older drug-adjacent options in dementia, high-dose vitamin E has a small evidence base at one stage.

1. Substance + claimed effects

Anti-amyloid monoclonal antibody infusions — currently lecanemab (Leqembi, Eisai/Biogen) and donanemab (Kisunla, Eli Lilly), both IgG1 humanised antibodies — are intravenous biologic therapies for early symptomatic Alzheimer's disease, meaning mild cognitive impairment due to AD or mild AD dementia with biomarker confirmation of cerebral amyloid pathology. Lecanemab binds preferentially to soluble amyloid-β protofibrils; donanemab binds an N-truncated pyroglutamate form of amyloid-β found specifically in established plaque van Dyck 2023 Sims 2023. The mechanistic claim is plaque clearance; the clinical claim is a 27–35% slowing of cognitive and functional decline over 18 months, measured on the CDR-SB and iADRS scales, in patients selected by amyloid PET or CSF biomarkers. The headline trade-off is amyloid-related imaging abnormalities (ARIA) — brain swelling (ARIA-E) and microhaemorrhages or superficial siderosis (ARIA-H) — that occur in roughly one in five lecanemab patients and one in three donanemab patients, with the risk concentrated in APOE-ε4 homozygotes van Dyck 2023 Sims 2023. Aducanumab (Aduhelm), the first FDA-approved drug in the class, was discontinued by Biogen in January 2024 and is no longer part of the active therapeutic landscape.

The article will cover this substance and its consequences holistically: cognitive trajectory in the treated patient (focus/mood-adjacent), short-term health impact via ARIA risk, longevity (modest, debated), the financial and effort burden of an 18-month infusion course with serial MRIs, eligibility gates, and the active controversy over whether the measured slowing is clinically meaningful. The brief named cognitive decline trajectory, plaque burden, ARIA, infusion logistics, and eligibility — all are in scope.

2. Evidence by addressing question

Mechanism

The amyloid cascade hypothesis posits that aggregated amyloid-β (Aβ) — soluble oligomers, protofibrils, and fibrillar plaques — is upstream of tau pathology, synaptic loss, and clinical dementia in AD. Anti-amyloid mAbs engage different conformers: aducanumab and lecanemab favour aggregated soluble forms (lecanemab with strong preference for protofibrils), donanemab targets a pyroglutamate-modified Aβ epitope concentrated in deposited plaque van Dyck 2023 Sims 2023. Binding triggers microglial phagocytosis (Fc-mediated) and direct disaggregation, both verified by amyloid PET reduction. In CLARITY-AD, lecanemab reduced amyloid PET centiloids from a mean baseline of ~77.9 by 55.5 centiloids at 18 months (placebo: +3.6) van Dyck 2023. In TRAILBLAZER-ALZ 2, donanemab reduced amyloid plaque by an average of 84% from baseline, with about half of participants reaching the pre-specified clearance threshold (<24.1 centiloids) by 12 months — at which point donanemab can be stopped, the only drug in the class with a finite-duration regimen tied to a biomarker endpoint Sims 2023. The mechanistic gap: amyloid is necessary but not sufficient for clinical AD. Tau, vascular pathology, and inflammation contribute independently, which is one reason the clinical effect (~25% slowing) is much smaller than the biological effect (>80% plaque reduction) Alves 2023.

Evidence

Two phase 3 RCTs anchor the evidence base, both biomarker-confirmed populations of mild cognitive impairment or mild AD dementia.

CLARITY-AD (lecanemab): 1,795 participants, 10 mg/kg IV every 2 weeks vs placebo, 18 months. Primary endpoint CDR-SB change from baseline: 1.21 (lecanemab) vs 1.66 (placebo), adjusted mean difference −0.45 (95% CI −0.67 to −0.23, p<0.001), a relative slowing of 27%. Secondary endpoints (ADAS-Cog14, ADCOMS, ADCS-MCI-ADL) all favoured lecanemab van Dyck 2023.
TRAILBLAZER-ALZ 2 (donanemab): 1,736 participants, IV monthly (titrated 700 mg ×3 then 1400 mg) for up to 72 weeks, with stopping when amyloid cleared. Primary endpoint iADRS change in low/medium-tau population: −6.02 (donanemab) vs −9.27 (placebo), absolute difference 3.25 points, 35% slowing; CDR-SB difference −0.67 (29% slowing). 39% reduction in risk of progressing to the next clinical stage. Combined population (including high-tau): 22% iADRS slowing, 29% CDR-SB slowing Sims 2023.

FDA traditional approval: lecanemab July 6, 2023; donanemab July 2, 2024 FDA 2023 FDA 2024. EMA: lecanemab approved 2024 with restrictions (excluding APOE-ε4 homozygotes); donanemab refused March 2025 on benefit-risk grounds EMA 2025. Aducanumab (Aduhelm), approved by FDA in 2021 under accelerated approval on amyloid-clearance grounds but with contested clinical data, was discontinued by Biogen January 2024 — not for safety/efficacy, but commercial reprioritisation toward lecanemab.

The clinical-meaningfulness controversy: the lecanemab −0.45 CDR-SB difference is below the 1-point minimum-clinically-important-difference threshold commonly cited for early AD; meta-analysis pooling lecanemab, donanemab, aducanumab, and solanezumab finds a class-level effect of −0.24 CDR-SB points, below all standard MCID thresholds Alves 2023. Proponents counter that 18 months is a short observation window and the divergence widens with time (open-label extensions at 36 months show continued separation) Petersen 2023.

Protocol

Lecanemab: 10 mg/kg IV over ~1 hour, every 2 weeks, indefinitely (no stop rule). Subcutaneous formulation (Leqembi IQLIK) approved 2025 for self-administered weekly maintenance after initiation phase. Donanemab: IV over 30 minutes, monthly, with titration (700 mg ×3 then 1400 mg). Donanemab is stopped when amyloid PET shows clearance (<24.1 centiloids on two consecutive scans or one scan <11 centiloids), achieved by ~50% of patients at 12 months and ~70% at 18 months Sims 2023. Both require MRI at baseline and serial monitoring MRIs: lecanemab before infusions 5, 7, and 14 (and per AUR, additional at 12 months for APOE4 carriers); donanemab before infusions 2, 3, 4, and 7 Cummings 2023. Modified-titration donanemab (TRAILBLAZER-ALZ 6, FDA-approved July 2025 label update): a slower 1-2-3-4 dose escalation cuts ARIA-E from 24% to 14% with preserved amyloid clearance, and from 57% to 19% in APOE4 homozygotes Wang 2025.

Contraindications

Per FDA labels and Appropriate Use Recommendations Cummings 2023:

More than 4 cerebral microbleeds, any area of superficial siderosis, prior macrohaemorrhage on baseline MRI
Anticoagulation (warfarin, DOACs) — relative contraindication; absolute for active treatment
Active or symptomatic ARIA — pause until radiographic resolution
Acute thrombolytic therapy (tPA, alteplase) — risk of catastrophic haemorrhage; documented fatal case in a lecanemab-treated patient who received tPA for ischaemic stroke Reish 2023
Cerebral amyloid angiopathy with prior haemorrhage
APOE-ε4 homozygotes: not a contraindication per FDA but excluded from EMA-approved lecanemab population due to disproportionate ARIA risk; AUR recommends APOE genotyping and informed-consent discussion of higher risk EMA 2025

ARIA-E (vasogenic oedema) and ARIA-H (microhaemorrhage / superficial siderosis) are the dose-limiting toxicity. CLARITY-AD: ARIA-E 12.6% (placebo 1.7%), ARIA-H 17.3% (placebo 9.0%); symptomatic ARIA-E 2.8%. TRAILBLAZER-ALZ 2: ARIA-E 24.0%, ARIA-H 31.4%; serious ARIA 1.6%; 3 deaths in donanemab arm attributable to ARIA van Dyck 2023 Sims 2023. APOE-ε4 homozygotes carry the steepest risk: ARIA-E 32.6% (lecanemab) and 40.6% (donanemab), with even higher rates for the standard titration in donanemab homozygotes (57% before modified titration) Hampel 2023 Wang 2025.

Misconceptions

Three persistent misconceptions in lay coverage:

"It reverses Alzheimer's." No. It slows decline. Patients still decline; the trajectory is shallower. Over 18 months, the absolute difference is about 5 months of preserved function on the CDR-SB scale — the treated 65-year-old in the trial reached the same disability state at roughly month 22 that the placebo group reached at month 18.
"It works for any dementia." No. It is approved only for biomarker-confirmed early-stage Alzheimer's disease. It is not indicated and not studied in vascular dementia, frontotemporal dementia, Lewy body dementia, or moderate-to-severe AD. Patients without amyloid pathology cannot benefit and still carry ARIA risk.
"Amyloid clearance equals cognitive cure." No. Plaque removal of >80% produces only ~25% slowing of clinical decline. Tau pathology, synaptic damage, and vascular contributions persist regardless. The amyloid cascade hypothesis is partial, not total.

Practicalities

Cost. Lecanemab list price: $26,500/year (US); typical Medicare beneficiary out-of-pocket ~$5,300/year before supplemental coverage CMS 2023. Donanemab: $32,000 for the full course (often ~12 months, capped because of the stopping rule). Total system cost including infusion administration, baseline and serial MRIs, amyloid PET (~$5,000–$7,000 if not covered), and APOE genotyping pushes the first-year burden above $40,000 per patient. CMS projected Medicare spending of $3.5 billion in 2025 for lecanemab alone.

Access. CMS requires participation in a clinical-evidence registry (the New IDEAS-adjacent CED structure) for Medicare coverage CMS 2023. Real-world rollout has been bottlenecked by amyloid PET capacity, infusion-centre availability, and neurologist throughput for the diagnostic workup. UK NICE rejected lecanemab in 2024 on cost-effectiveness grounds despite MHRA approval, so it is not available on the NHS.

Logistics. Patient time burden: ~26 infusions/year (lecanemab) at 1–2 hours/visit including pre-medication and post-infusion observation, plus 4–7 MRI scans in the first 14 months. The diagnostic pre-work — amyloid PET or lumbar puncture, APOE genotype, baseline MRI, cognitive testing battery — typically takes 6–12 weeks before first dose. Subcutaneous lecanemab maintenance dosing (approved 2025) reduces in-clinic time after the initiation phase.

Audience

Eligibility is narrow. Roughly 6.9 million Americans have AD; only ~10–15% are in the early symptomatic stage where these drugs are indicated. Of those, a substantial fraction fail biomarker confirmation, fail MRI screening (microbleeds), or have contraindications. Realistic eligible population in the US: 1–1.5 million. Patients are typically aged 60–85, MMSE 22–30, with mild functional impairment. APOE genotyping increasingly stratifies the decision: non-carriers face roughly halved ARIA risk; homozygotes (~15% of the eligible pool) face the steepest risk and EMA exclusion.

Alternatives

No directly comparable disease-modifying alternative exists. The class itself has lecanemab and donanemab as the only currently marketed agents. Symptom-modifying drugs (cholinesterase inhibitors — donepezil, rivastigmine, galantamine; NMDA antagonist memantine) offer modest, transient symptomatic benefit without altering trajectory. Trial-stage alternatives: anti-tau antibodies (early-stage), small-molecule amyloid inhibitors, lifestyle/exercise interventions (FINGER-style multidomain trials with ~25% cognitive improvement in at-risk populations, no biomarker effect). Aducanumab is withdrawn.

Failure-modes

The common patterns of "treatment failed":

Started too late: moderate or severe AD; outside trial population; no benefit expected.
Not biomarker-confirmed: clinical AD misdiagnosis; ~10–30% of "clinical AD" diagnoses lack amyloid pathology and cannot respond to anti-amyloid therapy.
Stopped early for asymptomatic ARIA: small ARIA detected on surveillance MRI, treatment paused; many patients restart but some discontinue, losing potential benefit.
Disease continues despite plaque clearance — the 22–35% slowing is the realistic ceiling; expectation of stabilisation or reversal sets up disappointment.
Caregiver burnout from infusion logistics (weekly to biweekly transport, MRI appointments, observation periods).

Stakes / payoff

For a typical 70-year-old with mild AD: untreated, expected progression to moderate dementia over 3–5 years with loss of independent ADLs and driving. With lecanemab over 18 months, the absolute delay on a global function scale is ~5 months — meaning, on the population-average trajectory, the milestones (forgetting an appointment, needing help with finances, getting lost driving home) arrive ~5 months later than they otherwise would. Whether that delay is felt as a meaningful gift of preserved time or as a marginal statistical artifact depends heavily on the patient and family.

3. Credibility range

Optimist case

Anti-amyloid mAbs are the first treatments in AD history to demonstrate disease modification — slowing progression on a clinical scale, not just masking symptoms. Replication is robust: lecanemab and donanemab independently show the same direction of effect with the same mechanism. The treatment-placebo curves diverge progressively in open-label extensions, suggesting cumulative benefit; modeling of CLARITY-AD extension data projects 2–3 years of delayed clinical milestones at 6 years. The 22–35% slowing is class-leading for any AD intervention. Donanemab's stop-rule design (treat to amyloid clearance, then stop) demonstrates the field has graduated to biomarker-guided regimens, with capped cost and exposure. ARIA is detectable, manageable, mostly asymptomatic, and resolves on treatment pause; APOE genotyping plus modified-titration dosing (donanemab) reduces ARIA-E by half in homozygotes. The validated amyloid cascade hypothesis enables a downstream pipeline (anti-tau, combination biologics) van Dyck 2023 Sims 2023 Petersen 2023.

Skeptic case

The clinical effect is statistically real but probably below the threshold of patient-noticeable benefit. The lecanemab −0.45 CDR-SB difference is half the most-cited 1-point MCID for early AD. Meta-analysis pooling all anti-amyloid mAb RCTs finds a class effect of −0.24 CDR-SB points Alves 2023. The 27% relative slowing translates to ~5 months of delayed progression over 18 months — a number the patient and family are unlikely to detect against the background variability of dementia. Meanwhile the harm is real and concrete: 21% ARIA on lecanemab, 36% on donanemab; ARIA-E with symptoms occurs in ~3% of patients; multiple deaths attributable to ARIA-related haemorrhage in trials and the open-label extension; a documented fatal interaction with tPA for stroke (a common comorbid event in this age group) Reish 2023. The cost burden ($26,500–$32,000/year drug, plus ~$15,000 in workup and monitoring) and operational lift (biweekly infusions, frequent MRIs, registry enrolment) consume opportunity-cost dollars from caregivers, families, and the Medicare programme. EMA's initial rejection of lecanemab and outright rejection of donanemab on benefit-risk grounds reflects a defensible read of the same data EMA 2025. The amyloid cascade hypothesis itself remains contested at the level of "necessary upstream cause" vs "downstream marker"; profound plaque clearance with marginal clinical benefit is evidence the hypothesis is incomplete.

Author's call

This is a real, FDA-approved, mechanism-validated disease-modifying therapy with a small clinical effect, significant safety burden, and a cost profile that creates genuine system-level tradeoffs. It is not a breakthrough; it is the first foothold. For the right patient — early-stage, biomarker-confirmed, APOE non-carrier or informed-consent heterozygote, no microbleeds, no anticoagulation, willing to commit to infusion and MRI logistics — the expected benefit-risk balance is positive but modest, and the decision is genuinely values-dependent. For the wrong patient (moderate AD, ε4-homozygote weighing the EMA-cited risks, anticoagulated, frail, rural without infusion access) the calculus tilts toward harm. The right action token is decide not do: this is a clinician-mediated choice with real downside, not a self-administered intervention to pursue. Evidence quality is high (large RCTs, FDA traditional approval, replication across two molecules); controversy is high (EMA-FDA divergence, ongoing MCID debate, class meta-analyses).

4. Stakeholder + incentive map

Manufacturers — Eisai/Biogen (lecanemab, $1.7B revenue projection 2025), Eli Lilly (donanemab). Commercial incentive to maximise eligible-population framing, downplay MCID critique, push subcutaneous/long-acting formulations to expand reach.
Alzheimer's Association — patient-advocacy body that endorsed FDA approval and CMS coverage early; funding ties to manufacturers documented.
FDA — granted accelerated approval to aducanumab over its advisory committee's negative vote (committee resignations followed); subsequent traditional approvals for lecanemab and donanemab on confirmatory phase 3 data.
CMS — coverage with evidence development requires registry participation; manages multi-billion-dollar projected spend.
EMA, NICE — more conservative regulators; EMA approved lecanemab (excluding ε4-homozygotes) and rejected donanemab; NICE rejected lecanemab as not cost-effective.
Skeptic camp — academic neurologists and methodologists (Alves, Karlawish, Whitehouse, Espay) publishing on clinical meaningfulness, conflict-of-interest disclosures, and the amyloid hypothesis itself. Public Citizen and consumer-watchdog groups oppose coverage expansion.
Neurology / memory-clinic infrastructure — beneficiary of the new patient flow (workup, infusion, monitoring revenue) but burdened by the operational lift; geographic access disparity is widening.

5. Population variability

Heterogeneity by genotype, sex, and pathology stage materially changes the calculus:

APOE-ε4 genotype. Roughly 15% of patients are ε4-homozygotes, 40% heterozygotes, 45% non-carriers. ARIA-E risk on lecanemab: 33% homozygotes, 11% heterozygotes, 5% non-carriers. ARIA-E risk on standard donanemab: 41% homozygotes, 23% heterozygotes, 15% non-carriers. Efficacy appears preserved or possibly enhanced in ε4 carriers Hampel 2023.
Sex. CLARITY-AD pre-specified subgroup analysis suggested larger effect in males than females, but the difference may be statistical artifact; pooled analyses are ongoing.
Tau burden. TRAILBLAZER-ALZ 2 stratified by tau PET. The low/medium-tau subgroup showed 35% iADRS slowing; the high-tau combined population fell to 22%, suggesting earlier intervention (less advanced disease) yields larger relative benefit Sims 2023.
Age and frailty. Trial populations centred on 60–85; over 85, frailty and competing mortality reduce expected net benefit. Under 65, atypical AD presentations may not meet inclusion biomarker criteria.
Race/ethnicity. Trial populations were ≥75% white. Black and Hispanic representation was 2–5% — too small for subgroup safety/efficacy analysis. AD prevalence is 2× higher in Black Americans; the data gap is significant.
Anticoagulation status. Approximately 20–30% of the eligible age cohort takes anticoagulation for atrial fibrillation or DVT prophylaxis. Coadministration with anti-amyloid mAbs substantially elevates haemorrhage risk; this is a hard exclusion in trials and a relative contraindication in practice.

6. Knowledge gaps

Long-term efficacy. Beyond 36 months, do the divergent curves continue to widen, plateau, or converge? Extension studies are ongoing but uncontrolled and selection-biased.
Pre-symptomatic treatment. AHEAD 3-45 (lecanemab in cognitively unimpaired amyloid-positive adults) and TRAILBLAZER-ALZ 3 (donanemab in preclinical AD) are testing whether earlier intervention yields larger benefit, but results not yet available.
Stopping criteria after clearance. Donanemab's stop rule is novel but not validated long-term: do plaques re-accumulate after stopping? Does cognitive benefit persist or fade? Initial extension data suggest re-accumulation is slow over 1–2 years.
Combination therapy. Anti-amyloid + anti-tau combinations are biologically attractive but not yet tested clinically.
Real-world ARIA rates. Trial populations exclude high-risk patients; community-practice ARIA rates may be higher.
MCID validation. The CDR-SB MCID for early AD is itself contested (proposed values range 0.5–2.0). Patient/caregiver-reported outcome measures may eventually settle whether the trial-measured slowing maps to lived experience.
Cost-effectiveness. ICER analyses have placed lecanemab's value-based price at $8,500–$20,600/year, well below list. The gap will likely widen the EMA-FDA-NICE regulatory divergence.

Scope as drafted. The brief named cognitive decline trajectory, plaque burden, ARIA, infusion logistics, and eligibility. All five are covered. The entry treats the class (currently lecanemab and donanemab) rather than picking a single molecule, because the active clinical question is class-level — both are FDA-approved, both face the same ARIA-and-cost trade, and a reader's neurologist will choose between them. Aducanumab is mentioned in the dossier but kept out of the article body since Biogen discontinued it in 2024.

Action token: decide, not do. Prescription-only, requires biomarker confirmation, narrow eligibility, real downside. Recommending it as a self-initiated do would mis-signal the burden of the call. The action label correctly puts this in the clinician-mediated decision bucket.

Cadence: course. Donanemab's stopping rule makes this most cleanly a course; lecanemab is technically ongoing but in practice many patients are treated for 18–36 months. course is the better fit than daily or weekly given the structured time-limited regimen and the bounded amyloid-clearance endpoint.

Rating notes.

focus = 2, longevity = 2. Genuinely a hard call. The trials show a real, replicated, statistically robust effect — but it's small in absolute terms (~5 months over 18) and below the most-cited MCID threshold. Scoring 3 felt like overclaiming given Alves 2023 and the EMA's read. Scoring 1 felt like dismissing two phase-3 RCTs and FDA traditional approval. 2 captures "real but small."
mood = 1. Indirect through preserved independence; no direct mood endpoint in either trial. Counterweighted by ARIA-monitoring anxiety. 1 captures the indirect signal honestly without inflating it.
health_short_term = 0. Within weeks the patient feels nothing positive; the felt experience is infusion logistics and MRI appointments. The cognitive endpoint moves over months, not weeks. Symptomatic ARIA is a short-term harm, not a benefit, so it sits in contraindications rather than this score.
cost_burden = 5, effort_burden = 4. 5 is correct for cost — drug + workup + monitoring exceeds the $10K/year prohibitive threshold easily, even with Medicare. Effort sits at 4 (an hour-plus per fortnight for infusions, plus serial MRIs and pre-work) rather than 5 because the regimen is structured and finite rather than dominating the whole day.
controversy = 4. EMA-FDA divergence and the MCID debate are foundational disagreements among credible camps. Not 5 (the field isn't a battleground in the homeopathy sense — both sides agree the drug clears amyloid and modestly slows decline; they disagree on whether that's worth it).

Audience scoping. Restricted to 60+ only because both trials enrolled patients ≥60 (TRAILBLAZER-ALZ 2: 60–85) or ≥50 (CLARITY-AD), but the practical eligible population is overwhelmingly in the 60+ band. Younger-onset AD patients are a small share, often with atypical presentations that may not meet biomarker criteria. Decision was to scope to the dominant treated population rather than over-inclusively scope to all adults.

Contraindication token. blood-thinners is the one that's most clearly load-bearing — the warfarin/DOAC interaction is a hard exclusion and the tPA case underscores it. Did not add cardiac-condition because not every cardiac patient is anticoagulated, and the contraindication is about the anticoagulation specifically. APOE-ε4 homozygote risk is real but doesn't map to any token in the closed vocabulary; handled in the audience section of the article instead.

Hard calls during the write.

Whether to lead the dek with the optimist or skeptic framing. Settled on optimist-then-honest: "first time in history" is true and the appropriate opening for a reference entry; the EMA dissent and cost trade follow inside the dek itself.
Whether to name lecanemab and donanemab explicitly in the title. Yes — the catalogue convention favours scientific name plus common-name disambiguation, and the parens form mirrors the UARS / ApoB precedents.
Whether to score energy/sleep nonzero (caregivers sleep worse around dementia onset, treatment may indirectly help). Settled on 0 — too indirect, not what the substance does, and adding it would dilute the meaningful scores.

Separate-entry candidates surfaced.

Amyloid PET / CSF biomarker testing for AD diagnosis — the diagnostic gate is substantial enough to warrant its own entry; covers indications, cost, sensitivity/specificity, blood-biomarker alternatives (p-tau 217 etc.).
APOE genotyping for late-onset AD risk — substantial enough to stand alone given growing direct-to-consumer testing and the ε4-homozygote-as-distinct-condition framing some researchers now use.
Cholinesterase inhibitors and memantine for AD symptom management — the symptom-management drug class deserves its own entry; not displaced by anti-amyloid therapy and often used alongside.
Multidomain dementia prevention (FINGER-style protocols) — pre-symptomatic prevention is a different decision in a different population and warrants its own entry rather than being squashed into alternatives here.

Future links. Once any of the above entries exist, they should be wired into the out-of-scope closing pointers. Also flag potential linkage to future entries on caregiver burden in dementia and advance care directives.

Knowledge gaps acknowledged but not surfaced to readers: long-term efficacy beyond 36 months, pre-symptomatic treatment (AHEAD 3-45, TRAILBLAZER-ALZ 3 ongoing), whether plaque re-accumulates after donanemab stopping, combination anti-amyloid + anti-tau therapy. These are in the research dossier but kept out of the article to avoid reader-facing speculation.