Therapy

1. Therapeutic Eras

For three decades after donepezil’s 1996 approval, every drug for Alzheimer’s disease was symptomatic: it raised acetylcholine or dampened glutamate, but left the underlying pathology untouched. Then in 2021, with the FDA accelerated approval of aducanumab, and definitively in 2023with lecanemab’s CLARITY-AD readout, the era of disease-modifying therapy began.

The shift is not subtle. For the first time, an AD drug shows both a molecular effect (amyloid PET goes negative) and a clinical effect (slower CDR-SB progression). The amyloid cascade hypothesis, first articulated by Hardy & Higgins in 1992 and covered in Part III, has finally produced a therapeutic.

2. Cholinesterase Inhibitors

The cholinergic hypothesis was articulated by Davies & Maloney (Lancet 1976): AD brain shows profound depletion of acetylcholine and severe loss of cholinergic neurons in the nucleus basalis of Meynert. If you cannot replace the lost neurons, you can at least slow the breakdown of the acetylcholine they release. Inhibit acetylcholinesterase (AChE), and synaptic ACh rises.

Effect size on cognition is modest: ~2.7 point improvement on the ADAS-Cog (70-point scale)versus placebo at 6 months — equivalent to delaying cognitive decline by ~2–4 months. Effects do not slow the underlying neurodegeneration; if you stop the drug, cognition rapidly returns to the untreated trajectory.

Pharmacokinetics — donepezil

The long half-life (~70 h) drives once-daily dosing and steady-state in ~2–3 weeks. For a one-compartment model:

\[ C_{ss} = \frac{F \cdot D}{\tau \cdot CL} \quad ; \quad t_{1/2} = \frac{\ln 2 \cdot V_d}{CL} \]

with F~100% oral bioavailability, V_d ~12 L/kg, CL ~0.13 L/h/kg. The high V_d reflects extensive tissue distribution; CYP2D6 and CYP3A4 mediate hepatic clearance, so poor metabolisers and inhibitor co-medications raise levels.

Adverse effects

• Cholinergic GI — nausea (~10%), vomiting, diarrhoea, anorexia, weight loss. Worst with rivastigmine PO; the patch halves GI burden.
• Bradycardia, syncope, AV-block — vagotonic effect; check ECG, beware co-prescribed β-blockers.
• Sleep disturbance & vivid dreams — donepezil at qhs.
• Muscle cramps, urinary urgency, increased secretions.
• Risk of cholinergic crisis in overdose or co-administration with other cholinergics.

3. Memantine

Memantine is a low-affinity, voltage-dependent, uncompetitive NMDA receptor antagonist. The hypothesis: chronic glutamatergic over-activation drives excitotoxic neurodegeneration, but high-affinity NMDA blockade (e.g. ketamine) is intolerable because it abolishes physiological signalling. Memantine’s rapid off-rate (k_off ≈ 1 s⁻¹) lets it block tonic excitotoxic currents while permitting fast synaptic transmission.

Approved 2003 (US) for moderate-to-severe AD(MMSE 3–14). The pivotal trial (Reisberg et al., NEJM 2003) showed modest improvement on CIBIC-Plus and ADCS-ADL.

Combination memantine + cholinesterase inhibitor (typically donepezil 10 mg + memantine 10 mg BID, packaged as Namzaric) is standard practice for moderate-severe AD; the additive effect is small but consistent across meta-analyses (Tariot et al., JAMA 2004).

4. The Amyloid-Targeting Antibody Era

The amyloid hypothesis (Part III) predicts: clear amyloid → halt the cascade. The translation took 25 years and several hundred billion dollars in failed trials. The arc of monoclonal antibody development — from the first immunotherapy attempt in 1999 to lecanemab’s positive readout in 2022 — encapsulates an entire generation of drug development.

Three lessons from the failures:

1. Target the right Aβ species. Solanezumab bound only soluble monomer; useless. Bapineuzumab bound everything but at low affinity. The successful antibodies (lecanemab, donanemab) target specific aggregated species — protofibrils and N-truncated plaque Aβ respectively.
2. Treat early. The earliest trials enrolled mild-to-moderate dementia, by which time tau pathology was self-propagating. CLARITY-AD and TRAILBLAZER-ALZ-2 enrolled MCI and mild dementia — the “early symptomatic” window.
3. Confirm amyloid biologically. Earlier trials enrolled clinical AD without biomarker confirmation; up to 25% were amyloid-PET negative and had no chance of benefit. Modern trials require amyloid PET or CSF Aβ42/40 ratio at screening.

All effective antibodies are humanised IgG1, leveraging FcγR-mediated microglial phagocytosis of opsonised plaques. None work without engaging microglia.

5. Aducanumab (Aduhelm)

Aducanumab was a fully human IgG1, derived from a B-cell clone in a healthy elderly donor with no cognitive symptoms (the “reverse translational” approach of Neurimmune; published as Sevigny et al., Nature 2016 — the “PRIME” phase 1b paper). It bound aggregated Aβ with strong selectivity over monomers (~10,000-fold).

EMERGE and ENGAGE

Two identical phase III trials (Biogen, run in parallel) were halted for futility in March 2019. A subsequent post-hoc re-analysis claimed that EMERGE (but not ENGAGE) showed a benefit at the high-dose arm — with the discrepancy attributed to slow titration in ENGAGE meaning fewer patients reached therapeutic exposure.

• EMERGE high dose: 22% slowing on CDR-SB at 78 weeks (p = 0.01)
• ENGAGE high dose: no benefit (p = 0.83); some signals of harm
• ARIA-E: 35% in 10 mg/kg arm (vs 3% placebo); ARIA-H 19%

The 2021 controversy

The FDA granted accelerated approval on 7 June 2021 on the basis of plaque clearance (a surrogate endpoint), against the advice of its own advisory committee (10–0 against, with one abstention). Three committee members resigned in protest, including Aaron Kesselheim. Medicare announced CMS would only pay for aducanumab in randomised clinical trials — effectively limiting access. Biogen withdrew the drug from the market in January 2024, ending a chapter that left the field with: (a) a shaken regulatory environment, (b) the first proof-of-concept that clearing plaque was clinically achievable, and (c) intense scrutiny of any successor.

6. Lecanemab (Leqembi) — flagship

Lecanemab (Eisai/BioArc/Biogen) is the humanised IgG1 version of the murine mAb158, raised against the “Arctic” APP mutation (E22G), which produces unusually stable Aβ protofibrils. It binds soluble protofibrils and short fibrils with ~1000-fold selectivity over monomer, and ~10-fold over insoluble fibrils — a binding profile distinct from aducanumab.

CLARITY-AD — the trial

van Dyck et al., NEJM 2023: double-blind, placebo-controlled, 1795 patients with early symptomatic AD (MCI or mild dementia, all amyloid-PET or CSF positive), 18 months. Lecanemab 10 mg/kg IV every 2 weeks vs placebo.

The 27% relative slowing translates to an absolute difference of 0.45 points on CDR-SB (range 0–18) at 18 months. Whether this is clinically meaningful is debated — some argue it equates to ~5 months of preserved function over the 18-month period; sceptics note it is below the typical “minimum clinically important difference” threshold (~1 point). The benefit is most apparent in patients with low baseline tau burden.

Mechanism — protofibril binding

Lecanemab’s binding profile predicts its mechanism: it scavenges the soluble oligomeric/protofibrillar species that are the most synaptotoxic forms of Aβ, while also engaging some plaque-associated protofibrils to drive microglial clearance. It binds the open conformation of the Aβ N-terminus, residues 1–6, with secondary contacts on residues 12–16.

Dosing and administration

• 10 mg/kg IV every 2 weeks; ~1-h infusion
• Indefinite duration — phase 3 OLE shows benefit accrues with continued dosing; protocol does not specify a stop point
• MRI at baseline, before infusions 5, 7 and 14, with additional scans for symptoms
• APOE genotyping required by FDA label since 2024 (for ARIA risk stratification)
• CMS coverage via QR-code real-world data registry

Pharmacokinetics: typical IgG1 — t_1/2 ~5–7 days, V_d~5 L (largely intravascular), clearance via FcRn-mediated recycling and target-mediated disposition.

\[ C_{ss,avg} = \frac{F \cdot D}{\tau \cdot CL} \approx \frac{1 \times 700\,\text{mg}}{14\,\text{d} \times 0.4\,\text{L/d}} \approx 125\,\mu\text{g/mL} \]

7. Donanemab (Kisunla) — flagship

Donanemab (Eli Lilly) is a humanised IgG1 directed against an N-truncated, pyroglutamate-modified form of Aβ(Aβ_p3-42), found almost exclusively within mature plaque cores. This narrow specificity — recognising a post-translationally modified epitope absent from soluble monomers — means donanemab essentially does not engage circulating Aβ; it only opsonises plaques. The result is an unusually aggressive plaque-clearing profile.

TRAILBLAZER-ALZ-2 — the trial

Sims et al., JAMA 2023: double-blind, placebo-controlled, 1736 patients with early symptomatic AD (MCI or mild dementia, all amyloid-PET positive AND tau-PET positive), 76 weeks. Patients were stratified by tau burden into “low/medium tau” (the prespecified primary analysis population, n = 1182) and “high tau” (advanced pathology).

The two highlights:

• Time-limited dosing. Once amyloid PET fell below the threshold (~11 centiloids) on two consecutive scans, donanemab dosing was switched to placebo. By week 76, ~62% of low/medium tau and ~30% of all participants had completed dosing — a feature unique among anti-amyloid antibodies.
• Tau stratification matters. The high-tau subgroup showed minimal benefit (presumed: tau pathology has decoupled from amyloid). This validated tau-PET as a clinical decision tool.

ARIA in donanemab

Donanemab’s ARIA rates are higher than lecanemab’s, reflecting more aggressive plaque engagement:

Dosing & administration

• Induction: 700 mg IV q4 weeks × 3
• Maintenance: 1400 mg IV q4 weeks until amyloid PET clears (typically 12–18 months)
• MRI at baseline, before infusion 2, 3, 4, and 7 (more than lecanemab, due to higher ARIA risk)
• APOE genotyping required by FDA label

8. ARIA — Amyloid-Related Imaging Abnormalities — flagship

ARIA is the dominant adverse-effect category of anti-amyloid antibodies, defining both the safety monitoring requirements and the limits of clinical use. Two forms, both visible on MRI:

Pathophysiology

The mechanistic model: anti-Aβ antibodies engage vascular amyloid in cerebral amyloid angiopathy (CAA) — the deposition of Aβ (mostly Aβ₄₀) in leptomeningeal and cortical vessel walls. Antibody binding triggers FcγR-mediated microglial activation along the vessel wall, transient breakdown of vascular integrity, and leak of plasma into the perivascular space (ARIA-E) or frank microhaemorrhage (ARIA-H). The risk is therefore proportional to baseline CAA burden — which is what APOE-ε4 dose primarily indexes.

APOE-ε4 dose dependence

Monitoring & management

• Pre-treatment MRI — exclude >4 microhaemorrhages, any superficial siderosis, prior macrohaemorrhage, or severe CAA on SWI.
• Surveillance MRI at 5, 7, and 14 (lecanemab) or before doses 2, 3, 4, 7 (donanemab); always when new neurological symptoms.
• Mild radiographic-only ARIA — continue dosing, repeat MRI in 1 month.
• Moderate or symptomatic ARIA — suspend, repeat MRI, consider corticosteroids if severe symptoms (the “ARIA storm” cases).
• Severe ARIA — permanent discontinuation; manage as CAA-related inflammation with high-dose steroids.
• Avoid t-PA and anticoagulation in any patient on or recently on these antibodies; anticoagulant exposure during ARIA has been linked to fatal haemorrhage (the Castellanos NEJM 2024 case report).

The pharmacokinetic basis for ARIA risk follows a threshold-exposure model:

\[ P(\text{ARIA-E}) \;=\; \Phi\!\left(\beta_0 + \beta_1 \cdot n_{\epsilon4} + \beta_2 \cdot \log C_{trough} + \beta_3 \cdot \text{CAA}_{baseline}\right) \]

where n_ε4 is APOE-ε4 allele dose (0/1/2), C_trough is steady-state antibody concentration, and CAA_baselineis microhaemorrhage count on screening MRI — the three dominant predictors.

9. Anti-Tau Therapy

Tau pathology correlates with cognitive decline far better than amyloid (Part IV), so an anti-tau therapeutic is the obvious next ambition. The pipeline has had a brutal decade: most antibodies have failed for efficacy. The lessons mirror those from the early anti-amyloid era — target the right species, treat early.

The most exciting current asset is BIIB080 (Biogen/Ionis), an intrathecal antisense oligonucleotide that suppresses MAPT (tau) mRNA translation. Phase I (Mummery et al., Nature Med 2023) showed ~55% reduction in CSF total tau over 6 months; Phase II readout expected 2026. A successful ASO would be the first true tau-modifying drug.

Why anti-tau antibodies have struggled

• Intracellular target. Most pathological tau is inside neurons; antibodies access mainly extracellular tau seeds during cell-to-cell propagation. The window is narrow.
• Heterogeneous strains. Tau in AD is a 3R+4R mix; PSP, CBD and FTD have different conformers. An antibody may fit one strain badly.
• Late treatment. By the time tau-PET is positive in inferior temporal cortex, neurons are dying.
• Targeting the N-terminus — which is cleaved off in the pathological core — bound only the “wrong end” of pathological tau. Mid-domain antibodies (bepranemab) and phospho-specific ones (JNJ-63733657) are showing slightly more promise.

10. Other Mechanism Approaches

Anti-inflammatory / microglial modulators

• Sodium oligomannate (GV-971) — approved in China 2019 for mild-moderate AD; proposed mechanism is gut microbiome modulation altering peripheral inflammation. Phase III GREEN MEMORY in Western populations halted 2022 for futility. Validity remains contested.
• TREM2 agonist antibodies — AL002 (Alector/AbbVie) targets the microglial TREM2 receptor implicated by GWAS (Part V). Phase II INVOKE-2 readout 2024 was negative; biology remains attractive.
• JAK inhibitors, masitinib, etanercept — all explored; nothing approved.

GLP-1 agonists

The most exciting non-amyloid axis. Liraglutide and semaglutide cross the blood-brain barrier modestly and have shown neuroprotection in models. The EVOKE/EVOKE-PLUS phase 3 trials of semaglutide 14 mg PO daily in early AD will read out 2026; the rationale combines insulin signalling, neuroinflammation, and the well-documented diabetes–dementia link. A positive readout would open an entirely new therapeutic axis.

Repurposing

• Lithium — long-standing observational link with reduced dementia incidence; small trials have shown CSF p-tau reduction at “microdose” (300 μg) levels (Forlenza). Definitive trial pending.
• Metformin — mixed observational data; the MASTER and AMPLIFY trials are testing prevention.
• SGLT2 inhibitors — emerging observational signal; no AD trials yet.
• Sildenafil — observational signal in EHR studies (Fang et al., Nat Aging 2021); causal claim disputed.
• Fluvoxamine, statins, valaciclovir, leflunomide — various ongoing trials based on observational and mechanistic signals.

BACE inhibitors — the cautionary tale

BACE1 inhibition cleanly reduces Aβ production; a generation of inhibitors (verubecestat, lanabecestat, atabecestat, elenbecestat, umibecestat) all failed phase III, several with cognitive worsening. BACE1 has unsuspected substrates (neuregulin-1, Sez6) whose inhibition produces dose-dependent cognitive impairment. The class is essentially abandoned.

11. Non-Pharmacologic Therapy

Lifestyle and environmental interventions have effect sizes that — especially cumulatively — rival the modern pharmacologic options, and far exceed them in cost-effectiveness. The Lancet Commission’s 14-factor model (Part I) estimates ~45% of dementia cases are theoretically preventable.

• Hearing aids — ACHIEVE trial (Lancet 2023). 3-year RCT in older adults with mild-to-moderate hearing loss: hearing-aid intervention plus best-practice audiologic care produced a 48% reduction in 3-year cognitive declinein the prespecified at-risk subgroup vs health-education control. Effect size comparable to lecanemab in absolute terms; effectively zero ARIA risk.
• Physical exercise. Aerobic + resistance training; ~150 min/week moderate intensity. Exerkines (irisin, BDNF, cathepsin B) cross the BBB and stimulate hippocampal neurogenesis. Multiple meta-analyses converge on ~20–30% relative risk reduction.
• Cognitive engagement. The FINGER trial (Lancet 2015; longer follow-up Stephen et al., JAMA Neurol 2024) was a multidomain intervention (diet + exercise + cognitive training + vascular risk monitoring) that showed a 25% benefit on a global cognitive composite; replicated by US-POINTER (2025).
• Mediterranean / MIND diet. Multiple cohorts; absolute benefit modest but consistent. Olive oil, leafy greens, fish.
• Sleep. Glymphatic clearance of Aβ is ~2× more efficient during NREM sleep (Xie et al., Science 2013); chronic short sleep predicts amyloid accumulation. Treat OSA, encourage sleep hygiene.
• Social engagement. Lonely older adults have ~50% greater dementia risk; the mechanism is unclear but the signal is replicable across cohorts.
• Vascular risk control. SPRINT-MIND (JAMA 2019): intensive BP control (target SBP < 120) reduced MCI by ~19% over ~5 years. This is the single most actionable midlife intervention.

12. Treatment Algorithm in 2026

The integration of biomarkers (Part VI) with the new disease-modifying therapies has changed the diagnostic-therapeutic workflow more in 3 years than in the preceding 30. A current algorithm:

1. Cognitive concern in patient >55.History, examination, MMSE/MoCA, depression screen, B12/TSH/blood count, structural MRI to exclude reversible causes.
2. If MCI or mild dementia — biomarker confirmation. Plasma p-tau217 is the new first-line screen (Barthélemy et al., Nat Med 2024; sensitivity ~95% for amyloid-PET positivity). Confirmatory amyloid PET or CSF Aβ42/40 + p-tau181.
3. If amyloid-positive and CDR-Global < 2:offer disease-modifying therapy. Required prerequisites:
   • APOE genotyping (counsel re ARIA, especially if ε4/ε4)
   • MRI screen for >4 microhaemorrhages, siderosis, prior macrohaemorrhage
   • Anticoagulation review — avoid concurrent therapy where possible
   • Reliable infusion-clinic access and surveillance MRI capacity
4. Choice of antibody. Lecanemab if access to q2-week infusion and lower-tau or non-ε4 patient; donanemab if patient values time-limited dosing and accepts higher ARIA risk. Both require continuous MRI surveillance.
5. Symptomatic therapy in parallel. Donepezil 10 mg qhs for MCI/mild AD (or rivastigmine patch if PO intolerant). Add memantine 10 mg BID once moderate (CDR ≥ 2 or MMSE < 15).
6. Lifestyle interventions universally.ACHIEVE-style hearing-aid optimisation, exercise prescription, BP control, sleep evaluation, social-engagement plan.
7. Stop antibody when: amyloid-PET clears (donanemab); ARIA serious or symptomatic; CDR-Global progresses to severe dementia (further benefit limited); or patient preference.
8. Decline anti-amyloid therapy if:amyloid-PET negative; CDR >= 2 with established advanced disease; high baseline CAA burden; on chronic anticoagulation that cannot be paused; geographic inability to perform surveillance MRI.

Who pays?

Lecanemab is priced at ~$26,500/year (US); donanemab at a similar level. Medicare covers both with the Coverage with Evidence Development real-world registry requirement. The infusion-clinic infrastructure cost (chair time, MRI capacity, specialist follow-up) substantially exceeds the drug cost in many systems and is the rate-limiting step on access.