Skip to content

Anti-amyloid immunotherapy

TL;DR — Two monoclonal antibodies have positive phase 3 primary endpoints. Lecanemab slowed 18-month CDR-SB decline by 0.45 points (95% CI 0.23–0.67; 1.21 vs 1.66 from a baseline of ~3.2) with ARIA-E in 12.6% and infusion reactions in 26.4% (van Dyck 2023, PMID 36449413). Donanemab slowed 76-week iADRS decline by 3.25 points (95% CI 1.88–4.62) in the low/medium-tau population and CDR-SB by 0.67 points (0.40–0.95), with ARIA-E in 24.0% (52 symptomatic) and three treatment-related deaths (Sims 2023, PMID 37459141). Aducanumab's two identical trials disagreed — EMERGE −0.39 CDR-SB (95% CI −0.69 to −0.09, P=0.012), ENGAGE +0.03 (−0.26 to 0.33, P=0.833) — and gantenerumab removed 56–66 Centiloids without slowing decline (−0.31 and −0.19 CDR-SB, both non-significant), while solanezumab, which targets monomers rather than aggregates, failed in preclinical AD over 240 weeks (Budd Haeberlein 2022, PMID 35542991; Bateman 2023, PMID 37966285; Sperling 2023, PMID 37458272). The effect sizes that succeeded are smaller than published minimal clinically important differences for the same instruments — +1 to +2 CDR-SB points depending on stage (Muir 2024, PMID 38561021) — which is the substance of the clinical-meaningfulness dispute, not a rhetorical position. Real-world safety is workable but not trivial: in 234 consecutively treated patients, 22% developed ARIA, 5.7% symptomatic ARIA, and 4.3% withdrew for ARIA, with symptomatic ARIA in 27% of those with mild dementia versus 1.8% of those with MCI or very mild dementia (Paczynski 2025, PMID 40354064).

The trials

Agent Trial(s) n Duration Primary result Amyloid change ARIA-E Notable harm
Lecanemab (protofibril-directed) CLARITY AD (NCT03887455) 1,795 18 mo CDR-SB 1.21 vs 1.66; difference −0.45 (95% CI −0.67 to −0.23), P<0.001 −59.1 Centiloids (95% CI −62.6 to −55.6) in 698-participant substudy 12.6% Infusion reactions 26.4% (van Dyck 2023, PMID 36449413)
Donanemab (pyroglutamate-Aβ) TRAILBLAZER-ALZ 2 (NCT04437511) 1,736 76 wk iADRS −6.02 vs −9.27; difference 3.25 (1.88–4.62), P<0.001 in low/medium tau; combined population 2.92 (1.51–4.33). CDR-SB difference −0.67 (−0.95 to −0.40) low/medium tau; −0.70 (−0.95 to −0.45) combined. 23 of 24 gated outcomes significant Not reported in the primary abstract 24.0% (52 symptomatic) 3 treatment-related deaths (vs 1 placebo); infusion reactions 8.7% (Sims 2023, PMID 37459141)
Aducanumab EMERGE + ENGAGE 1,638 + 1,647 78 wk EMERGE high dose −0.39 (−0.69 to −0.09), P=0.012, 22% decrease; ENGAGE +0.03 (−0.26 to 0.33), P=0.833 Dose- and time-dependent biomarker reduction in both Most common AE Both trials halted for futility on ~50% of enrolees; only 55.2% completed (Budd Haeberlein 2022, PMID 35542991)
Gantenerumab (subcutaneous) GRADUATE I + II (NCT03444870, NCT03443973) 985 + 980 116 wk CDR-SB difference −0.31 (−0.66 to 0.05), P=0.10 and −0.19 (−0.55 to 0.17), P=0.30 −66.44 and −56.46 Centiloids; amyloid-negative in 28.0% and 26.8% 24.9%; symptomatic 5.0% Lower CSF p-tau181 and higher Aβ42 vs placebo, but tau-PET accumulation similar (Bateman 2023, PMID 37966285)
Solanezumab (monomer-directed) EXPEDITION3 (NCT01900665), mild dementia 2,129 80 wk ADAS-cog14 6.65 vs 7.44; difference −0.80 (−1.73 to 0.14), P=0.10. MMSE −3.17 vs −3.66 (descriptive after hierarchical failure) Not a plaque-clearing antibody ARIA-E 1 vs 2 patients Failed at the mild-dementia stage before A4 tested the preclinical stage (Honig 2018, PMID 29365294)
Solanezumab (monomer-directed) A4 (NCT02008357), preclinical AD 1,169 240 wk PACC change −1.43 vs −1.13; difference −0.30 (−0.82 to 0.22), P=0.26 Amyloid increased in both arms: +11.6 vs +19.3 Centiloids <1% in both arms ARIA-H 29.2% vs 32.8% — i.e. background rate (Sperling 2023, PMID 37458272)
Donanemab (phase 2, TRAILBLAZER-ALZ) NCT03367403 257 76 wk iADRS −6.86 vs −10.06; difference 3.20 (0.12–6.27), P=0.04. Most secondary outcomes no substantial difference. Amyloid reduction 85.06 Centiloids greater than placebo; global tau load difference 0.01 Deep plaque clearance with almost no tau-PET movement ARIA-E (mostly asymptomatic) An iADRS effect of similar magnitude was already present in 257 patients, though not in the same population — phase 3's 3.25 is the low/medium-tau subgroup, phase 2's 3.20 the whole trial (Mintun 2021, PMID 33720637)

Read as a set, the pattern is consistent with a dose–response on plaque removal rather than with amyloid-antibody class failure: solanezumab did not remove plaque and did nothing; gantenerumab removed substantial plaque but rendered only ~27% amyloid-negative and did not slow decline significantly; lecanemab and donanemab produced deeper clearance and slowed decline. It is also consistent with a threshold model in which clearance must be near-complete to matter. It does not establish that amyloid removal per se is the mechanism of benefit — see amyloid biology.

Head-to-head clearance

TRAILBLAZER-ALZ 4 randomised 148 participants open-label to donanemab or aducanumab. Amyloid plaque clearance (<24.1 Centiloids) was achieved by 37.9%, 70.0% and 76.8% of donanemab-treated participants at 6, 12 and 18 months versus 1.6%, 24.6% and 43.1% with aducanumab (P<0.001); median time to clearance 359 versus 568 days (P<0.001). ARIA-E occurred in 23.9% and 34.8% respectively, and the authors note that the depth and speed of removal did not affect ARIA risk (Salloway 2025, PMID 40390253). This is a biomarker trial with no clinical endpoint.

Is the effect clinically meaningful?

The published MCIDs for AD trial endpoints are: MCI — ADAS-Cog +2 to +3, CDR-SB +1, iADRS −5, MMSE −1 to −2; mild AD — ADAS-Cog +3, CDR-SB +2, iADRS −9, MMSE −2 (Muir 2024, PMID 38561021). Against these:

Trial effect MCID for that stage Ratio
Lecanemab CDR-SB −0.45 +1 (MCI) to +2 (mild AD) 0.23–0.45×
Donanemab CDR-SB −0.67 +1 to +2 0.34–0.67×
Donanemab iADRS 3.25 −5 (MCI) to −9 (mild AD) 0.36–0.65×

The review's own conclusion is that average treatment effects in recent anti-amyloid trials are lower than previously published MCIDs, and that MCIDs were derived without systematic patient and care-partner input (PMID 38561021). Three positions are defensible on this evidence and are all held in the field: (a) a group-mean difference below an individual-level MCID can still be clinically important if some individuals benefit substantially; (b) MCIDs derived for cross-sectional interpretation are the wrong yardstick for a slowing-of-decline effect; (c) the effect is real but too small to justify the burden and cost. This knowledge base records the disagreement rather than resolving it; the design of a study that could resolve it is in OPEN-QUESTIONS.md.

ARIA

ARIA-E (oedema/effusion) and ARIA-H (microhaemorrhage/haemosiderosis) share clinical, biological and pathophysiological features with AD and cerebral amyloid angiopathy. Across randomised trials most ARIA cases were asymptomatic; symptomatic ARIA-E occurred more at higher doses and generally resolved within 3–4 months or on stopping treatment. APOE haplotype and dose are the major risk factors, and any microhaemorrhage on baseline MRI increases risk (Hampel 2023, PMID 37280110).

Mechanistically, autopsy of five aducanumab-treated patients found that regions corresponding to ARIA on MRI contained microinfarcts with haemosiderin, complement activation and CD68-positive vessel walls originating from Aβ-laden leptomeningeal and penetrating vessels, with Aβ clearance restricted to cortical layer I (Boon 2025, PMID 41109234). ARIA is best understood as iatrogenic destabilisation of vascular amyloid — see vascular and metabolic contributions.

Quantified risk

Source Population ARIA rate
Qi 2026 meta-analysis (PMID 41478817) 2 RCTs + 5 real-world studies, 1,576 lecanemab-treated patients Pooled ARIA 19% (95% CI 16–23); symptomatic ARIA 3% (2–4); infusion reactions 26% (19–34); discontinuation for AEs 8% (5–11); discontinuation for ARIA 5% (3–7) in real-world studies. APOE4 gene-dose effect: RR 1.45 (heterozygotes) and 3.54 (homozygotes) vs non-carriers
Paczynski 2025 (PMID 40354064) 234 consecutive patients, one US specialty memory clinic, Aug 2023–Oct 2024, mean 6.5 months of treatment Infusion reactions 37% (typically mild); ARIA 22% of 194 at risk; ARIA-E ± ARIA-H 15%; isolated ARIA-H 6.7%; symptomatic ARIA 5.7%, clinically severe 1.0%; no macrohaemorrhage and no deaths; 9.8% withdrew overall, 4.3% for ARIA. 23% had microhaemorrhage or siderosis before starting. Symptomatic ARIA 27% in mild dementia vs 1.8% in MCI/very mild dementia
Trial rates CLARITY AD / TRAILBLAZER-ALZ 2 / GRADUATE ARIA-E 12.6% / 24.0% / 24.9% (PMID 36449413; PMID 37459141; PMID 37966285)

The stage gradient in the real-world data — 27% symptomatic ARIA in mild dementia versus 1.8% in MCI — is the single most actionable number for treatment selection and is not visible in the pooled trial figures. The APOE4 gene-dose effect is the reason genotyping precedes treatment; note that APOE4 homozygotes are also the group with near-fully-penetrant AD biology and earliest onset (Fortea 2024, PMID 38710950), so the highest-risk group for ARIA overlaps with the group with the strongest indication.

Appropriate use, in practice

Element Lecanemab AUR (Cummings 2023, PMID 37357276) Donanemab AUR (Rabinovici 2025, PMID 40155270)
Stage MCI due to AD or mild AD dementia with confirmed amyloid pathology MCI or mild dementia due to AD, Clinical Stages 3–4, MMSE 20–30
Biomarker Confirmed brain amyloid pathology PET or CSF confirmation; tau PET not required for eligibility
APOE Genotyping recommended to inform risk discussion; ε4 carriers, especially homozygotes, at higher risk Genotyping should be performed before treatment
Baseline MRI Required; protocols for serious events must exist Within 12 months of starting; exclude if >4 cerebral microbleeds, cortical superficial siderosis, or major vascular contribution to cognitive impairment
Anticoagulation Recommends patients requiring anticoagulants not receive lecanemab until more data are available —
Monitoring Detailed ARIA and infusion-reaction monitoring Surveillance MRI before the 2nd, 3rd, 4th and 7th infusions; before the 12th dose in higher-risk individuals; and whenever ARIA is suspected
Stopping — May consider discontinuing when amyloid clearance is demonstrated on PET, typically 12–18 months after starting
Framing Clinician and institutional preparedness mandatory; communication and trust-building central Shared decision-making grounded in the patient's values and goals of care

Both documents deliberately track trial inclusion and exclusion criteria, on the explicit reasoning that safety and efficacy are known only for patients like trial participants (PMID 37357276).

Who is actually eligible?

Applying the AURs to 1,005 amyloid-positive attendees at a Korean memory clinic (from 2,726 total visitors), 24.6% were eligible for lecanemab and 28.0% for donanemab — 9.1% and 10.3% of the original clinic population. Using demographically adjusted MMSE z-score thresholds instead of raw MMSE raised eligibility to 38.4% and 33.7%. More than a third were excluded on MRI findings, mainly vascular lesions carrying ARIA risk (Jeon 2025, PMID 41225766).

The general point survives the single-country setting: after biomarker confirmation, roughly a quarter of amyloid-positive patients meet the recommendations, and the commonest disqualifier is cerebrovascular burden — which is also the commonest copathology. The population that qualifies is systematically less mixed-pathology than the population with dementia, which is the selection gap discussed on vascular and metabolic contributions.

Health-system readiness is a separate constraint: identifying eligible patients, delivering treatment and providing surveillance MRI at scale requires infrastructure that most systems do not have, and the shift from syndrome-based dementia care to biomarker-guided treatment demands substantial adjustment of resources (Heneka 2024, PMID 39549715). Delivery is changing: the FDA approved a subcutaneous lecanemab maintenance regimen in 2025 and an at-home subcutaneous starting regimen on 13 July 2026. That can reduce infusion-centre demand, but it does not remove amyloid confirmation, APOE-informed counselling, baseline/surveillance MRI or emergency pathways for suspected ARIA (FDA, https://www.fda.gov/drugs/news-events-human-drugs/fda-approves-first-home-starting-dose-alzheimers-disease-treatment, directly fetched 2026-08-31). See care, caregiving and health systems.

Cost and access

A modelled cost-effectiveness and threshold analysis across 174 countries estimated that lecanemab and donanemab increase average quality-adjusted life years by 0.38 and 0.51 respectively versus usual care, and derived value-based prices ranging from US$254–9,434 (lecanemab) and US$387–13,964 (donanemab) in high-income countries, US$90–1,025 and US$137–1,507 in upper-middle-income countries, US$11–623 and US$21–956 in lower-middle-income countries, and US$4–18 and US$9–32 in low-income countries (Hoang 2026, PMID 42125995). The four-order-of-magnitude spread is a statement about willingness-to-pay thresholds, not about biology — but combined with the projection that the largest growth in dementia burden falls in low- and middle-income regions (epidemiology and burden), it describes a therapy that is priced out of the places where the burden is growing fastest.

What is unsettled

What the extension data add—and cannot add

The CLARITY-AD open-label extension reports continued separation of early-start and delayed-start cohorts through 36 months and no new safety signal, but the post-randomisation period is uncontrolled and the abstract does not provide a new placebo-adjusted CDR-SB estimate; it supports durability of target engagement more strongly than a causal 36-month clinical effect (van Dyck 2025, PMID 41355080). The expanded safety population is more decision-relevant: among 1,612 people receiving at least one lecanemab dose across core plus extension, ARIA-H microhaemorrhage occurred in 16.0%, ARIA-E in 13.6%, and infusion reactions in 24.5%. ARIA-E reached 34.5% in APOE ε4 homozygotes. Two lecanemab recipients in the extension had concurrent fatal intracerebral haemorrhage—one after tissue plasminogen activator and one while anticoagulated—without proving causation but sharpening the exclusion logic used by appropriate-use recommendations (Honig 2024, PMID 38730496; NCT03887455).

The controversy is therefore asymmetric. The 18-month randomised effect is causal but small; the 36-month extension is longer but vulnerable to selection, attrition and absence of concurrent control. Conversely, rare catastrophic harms may become clearer only after more exposure. Long follow-up improves precision without recreating randomisation.

  1. Duration. Neither pivotal trial establishes what happens after 18 months, whether benefit accumulates, plateaus or reverses, or whether treatment can be stopped after clearance. The donanemab AUR's stopping guidance is a pragmatic inference, not trial evidence (PMID 40155270).
  2. Prevention. Solanezumab failed in preclinical AD, but it also failed to remove plaque (PMID 37458272); whether a clearing antibody prevents symptom onset is being tested — see clinical trials landscape.
  3. Mechanism of benefit. Gantenerumab produced large Centiloid reductions and lowered CSF p-tau181 without slowing decline or altering tau-PET accumulation (PMID 37966285), so amyloid reduction is not sufficient.
  4. Generalisability. Trials enrolled biomarker-confirmed, largely non-anticoagulated, low-microbleed participants; the treated population will drift from that as access widens.

Open questions

  • Is a group-mean CDR-SB difference of 0.45–0.67 points meaningful to patients and care partners, and what design would answer that — responder analyses, time-saved metrics, or preference-weighted outcomes (Muir 2024, PMID 38561021)?
  • Why did EMERGE and ENGAGE disagree despite identical protocols, and does the answer generalise to other agents (Budd Haeberlein 2022, PMID 35542991)?
  • Why did gantenerumab clear 56–66 Centiloids without clinical benefit while lecanemab cleared 59 Centiloids with benefit — depth of clearance, epitope, trial population, or chance (Bateman 2023, PMID 37966285; van Dyck 2023, PMID 36449413)?
  • Does the 27% symptomatic-ARIA rate in mild dementia mean treatment should be restricted to MCI (Paczynski 2025, PMID 40354064)?
  • Should APOE4 homozygotes — highest ARIA risk (RR 3.54) but earliest and most penetrant disease — be treated earlier, later, differently, or not at all (Qi 2026, PMID 41478817; Fortea 2024, PMID 38710950)?
  • Can treatment be stopped once amyloid clearance is achieved, and does pathology re-accumulate (Rabinovici 2025, PMID 40155270; Salloway 2025, PMID 40390253)?
  • Is anticoagulation an absolute or relative contraindication? The lecanemab AUR advises against treatment pending data that do not yet exist (Cummings 2023, PMID 37357276).
  • At value-based prices of US$4–18 per year in low-income countries, is any access model viable where the burden is growing fastest (Hoang 2026, PMID 42125995)?

References

  1. van Dyck CH, et al. Lecanemab in early Alzheimer's disease. N Engl J Med. 2023;388:9-21. PMID 36449413.
  2. Sims JR, et al. Donanemab in early symptomatic Alzheimer disease: the TRAILBLAZER-ALZ 2 randomized clinical trial. JAMA. 2023;330:512-527. PMID 37459141.
  3. Budd Haeberlein S, et al. Two randomized phase 3 studies of aducanumab in early Alzheimer's disease. J Prev Alzheimers Dis. 2022;9:197-210. PMID 35542991.
  4. Bateman RJ, et al. Two phase 3 trials of gantenerumab in early Alzheimer's disease. N Engl J Med. 2023;389:1862-1876. PMID 37966285.
  5. Sperling RA, et al. Trial of solanezumab in preclinical Alzheimer's disease. N Engl J Med. 2023;389:1096-1107. PMID 37458272.
  6. Salloway S, et al. TRAILBLAZER-ALZ 4: a phase 3 trial comparing donanemab with aducanumab on amyloid plaque clearance in early, symptomatic Alzheimer's disease. Alzheimers Dement. 2025;21:e70293. PMID 40390253.
  7. Hampel H, et al. Amyloid-related imaging abnormalities (ARIA): radiological, biological and clinical characteristics. Brain. 2023;146:4414-4424. PMID 37280110.
  8. Qi L, et al. Safety profiles of lecanemab: a systematic review and meta-analysis of randomized controlled trials and real-world evidence. J Prev Alzheimers Dis. 2026;13:100473. PMID 41478817.
  9. Paczynski M, et al. Lecanemab treatment in a specialty memory clinic. JAMA Neurol. 2025;82:655-665. PMID 40354064.
  10. Cummings J, et al. Lecanemab: appropriate use recommendations. J Prev Alzheimers Dis. 2023;10:362-377. PMID 37357276.
  11. Rabinovici GD, et al. Donanemab: appropriate use recommendations. J Prev Alzheimers Dis. 2025;12:100150. PMID 40155270.
  12. Jeon SY, et al. Eligibility for lecanemab and donanemab in Korea under appropriate use recommendations. Alzheimers Dement. 2025;21:e70875. PMID 41225766.
  13. Hoang MT, et al. Value-based prices of emerging disease-modifying therapies for Alzheimer's disease in 174 countries: a cost-effectiveness and threshold analysis. Alzheimers Dement. 2026;22:e71308. PMID 42125995.
  14. Heneka MT, et al. Passive anti-amyloid β immunotherapy in Alzheimer's disease — opportunities and challenges. Lancet. 2024;404:2198-2208. PMID 39549715.
  15. Muir RT, et al. Minimal clinically important difference in Alzheimer's disease: rapid review. Alzheimers Dement. 2024;20:3352-3363. PMID 38561021.
  16. Boon BDC, et al. Neuropathological changes and amyloid-related imaging abnormalities in Alzheimer's disease treated with aducanumab versus untreated: a retrospective case-control study. Lancet Neurol. 2025;24:931-944. PMID 41109234.
  17. Fortea J, et al. APOE4 homozygozity represents a distinct genetic form of Alzheimer's disease. Nat Med. 2024;30:1284-1291. PMID 38710950.
  18. Honig LS, et al. Trial of solanezumab for mild dementia due to Alzheimer's disease. N Engl J Med. 2018;378:321-330. PMID 29365294.
  19. Mintun MA, et al. Donanemab in early Alzheimer's disease. N Engl J Med. 2021;384:1691-1704. PMID 33720637.
  20. van Dyck CH, et al. Long-term safety and efficacy of lecanemab in early Alzheimer's disease: results from the CLARITY AD open-label extension study. Alzheimers Dement. 2025;21:e70905. PMID 41355080.
  21. Honig LS, et al. Updated safety results from phase 3 lecanemab study in early Alzheimer's disease. Alzheimers Res Ther. 2024;16:105. PMID 38730496.

Non-journal source. FDA. "FDA approves first at-home starting dose for Alzheimer's disease treatment." 13 July 2026. https://www.fda.gov/drugs/news-events-human-drugs/fda-approves-first-home-starting-dose-alzheimers-disease-treatment. Directly fetched 2026-08-31.