An Alzheimer's treatment that halves decline is approved

Bottom line

I estimate about a 26% chance of YES by the end of 2031, and close to zero before 2028. Almost all of that comes from a few large antibody trials that report between late 2027 and 2029. Approvals based on them could arrive between late 2028 and 2031.

Where things stand (as of 16 Oct 2026)

• No approved Alzheimer's drug meets the 50% bar.
  • Leqembi (lecanemab) slowed decline on its main scale (CDR-SB) by 27%.
  • Kisunla (donanemab) slowed decline by 22% in the overall population and 35% in the low/medium-tau group, both on its primary scale (iADRS). The low/medium-tau group was a primary population. (Lilly press release, 17 Jul 2023.)
• Nothing qualifying has been approved since the question opened. The subcutaneous starting dose of Leqembi was approved on 13 Jul 2026, before the window, and relies on the same 27% trial result.
• Several other candidates are out or weak:
  • Anavex withdrew its EU application for blarcamesine in March 2026.
  • Novo Nordisk's semaglutide trials (EVOKE) failed.
  • Biogen's diranersen missed its primary dose-response objective; its best dose showed 26% slowing on CDR-SB (Alzforum, Jul 2026).

Reference class

• About 20 Phase 3 trials of disease-modifying drugs have reported. None reached 50% in a primary analysis; the best was 35%.
• Subgroups at earlier disease stages look stronger:
  • Kisunla, mild cognitive impairment with low/medium tau: 60% slowing on iADRS and 46% on CDR-SB.
  • In Kisunla's main trial, the risk of moving to the next disease stage fell by 39% (hazard ratio 0.61).
  • In the inherited-Alzheimer's prevention study (DIAN-TU) of gantenerumab, the open-label extension gave a hazard ratio of 0.79 in people without symptoms, and 0.53 in those treated for about 8 years.
• Older symptom-relief drugs (donepezil and its class) did exceed 50%, because patients on the drug improved while placebo patients declined. No new symptom-relief drug has been approved since 2003.

Paths to YES (my modelled chance, including approval)

1. AHEAD 3-45, lecanemab, people without symptoms (about 12–14%)
   • The primary endpoint is change on a cognitive test battery (PACC5) at week 216. Primary completion is 21 Dec 2028 (ClinicalTrials.gov), with readout expected in 2028–29.
   • The trial was designed with 80% power to detect 38% slowing (BioArctic design poster).
   • In the A4 study, amyloid-negative volunteers in the companion LEARN study did not decline on this battery over 240 weeks. So most of the decline in the placebo group is disease-driven, and a ratio of 50% or more is reachable through a large true effect or sampling noise.
   • If it succeeds, approval would most likely come around 2030.
2. TRAILBLAZER-ALZ 3, donanemab, people without symptoms (about 5–6%)
   • The primary endpoint is time to clinical progression in participants with a baseline CDR global score of 0. The trial is event-driven (about 350 events); primary completion is Nov 2027 and readout is expected late 2027 (Baron Q3 2026 letter).
   • It is powered for a hazard ratio of about 0.7. In A4, 19%, 35% and 41% of participants progressed in the low, middle and high amyloid groups (Rentz et al., JPAD 2024). Progression unrelated to Alzheimer's dilutes the effect, so a hazard ratio of 0.5 or lower needs roughly 65–75% prevention of Alzheimer's-driven progression.
   • Lilly's chief scientific officer has said he expects "profound results."
3. TRONTIER 1 and 2, trontinemab, early symptomatic disease (about 6%)
   • Primary endpoint is CDR-SB at week 72; data are due in the first half of 2028 (Roche Pharma Day, 28 Sep 2026). Either of the two trials reaching 50% is enough.
4. TRAILRUNNER-ALZ 3, remternetug, people with no or minimal impairment (about 4–5%)
   • Primary completion is Apr 2029, so approval would fall around 2030–31.
5. Smaller paths (about 5% combined)
   • Buntanetap: 6-month readout in Q1 2027, 18-month readout in Q1 2028; a short trial where patients improve could clear 50%.
   • AR1001 and Xanamem: topline results due in November 2026.
   • Unknown candidates.
   • A residual chance that a resolver counts a reformulation of an older drug.

These paths overlap, because one hidden factor drives several of them: how well removing amyloid early works. I combined them in a Monte Carlo model:

• That efficacy factor was drawn from Beta priors with means of 0.31, 0.375 and 0.48, weighted 0.35, 0.45 and 0.20.
• Each trial has its own sampling noise, an approval probability of 0.70–0.85, and a review lag of 0.75–1.6 years.
• The ensemble gave 27% by the end of 2031. I shaded this to about 26% for historical over-optimism and possible delays.

How I read the question

A generic, or a reformulation that relies on donepezil-era trials, mostly fails two tests:

• the qualifying arm must use "a regimen covered by the approval";
• the pivotal trial must be identified in the regulator's own assessment.

I gave this route about 1%.

Strongest case against

• Every disease-modifying drug so far has fallen in the 20–35% range.
• The only completed prevention trial in people without symptoms (A4) failed.
• FDA may be wary of preclinical indications because of brain swelling and bleeding risks (ARIA).
• All three main trials could plausibly come in at 30–40%: meaningful, but NO for this question.

This is why I stay near 25% rather than 35–45%.

Sources

• ClinicalTrials.gov records: NCT05026866, NCT04468659, NCT06653153, NCT07169578.
• Lilly Q3 2025 earnings call (Motley Fool transcript).
• Alzforum: Nov 2024 coverage; Kisunla and Leqembi pages (Sep 2026).
• Yaari et al. 2025, PMC12439032.
• Roche Pharma Day deck, 28 Sep 2026.
• Baron Q3 2026 letter.
• Anavex, Annovis, AriBio and Actinogen releases, 2026.
• MedPage on DIAN-TU, Mar 2025.
• Sperling and Rentz, JPAD 2024.

Question. By the last day of each quarter, has the FDA, or the European Commission following EMA review, approved an Alzheimer's disease treatment whose pivotal trial showed at least a 50% relative reduction in clinical worsening versus its concurrent control in a prespecified primary efficacy analysis?

Definitions.

• Approval: an FDA marketing approval for an Alzheimer's disease indication, or a European Commission marketing authorisation for that indication following EMA review. Initial approvals and supplemental approvals or indication extensions count, including accelerated or conditional approvals if the clinical-efficacy requirement below is independently satisfied. A favourable EMA committee opinion alone is not approval.
• Treatment: a medicinal product or combination regimen approved to treat Alzheimer's disease, including mild cognitive impairment due to Alzheimer's disease. No AI involvement in discovery, development, or evaluation is required.
• Pivotal trial: a human controlled trial identified in the approving regulator's assessment as providing pivotal or confirmatory efficacy evidence for the approved indication. At least one such trial must qualify; not every pivotal trial must reach 50%. The qualifying treatment arm must use a regimen covered by the approval.
• Clinical worsening: deterioration in cognition, everyday functioning, or a global clinical measure of Alzheimer's disease severity, measured by a clinical scale such as CDR-SB, iADRS, ADAS-Cog, or ADCS-ADL, or by a defined clinical-progression event. Biomarker changes alone, including amyloid clearance, tau reduction, or imaging changes, do not qualify. Improvement in agitation, sleep, or another isolated associated symptom alone does not qualify.
• Prespecified primary efficacy analysis: the endpoint, analysis population, time point, comparison, and statistical method were designated as primary or co-primary in the protocol or statistical analysis plan before treatment assignments were unblinded for the relevant efficacy analysis. A biomarker-defined or other restricted population counts if it was designated a primary efficacy population and is included in the approved indication. A merely prespecified subgroup analysis of a primary endpoint does not count. One qualifying primary or co-primary analysis is sufficient.
• Concurrent control: participants enrolled contemporaneously in a separate control arm of the same trial, receiving placebo, usual care, or an active comparator. Historical controls, external natural-history cohorts, and modelled controls do not count.
• At least 50% relative reduction: the primary-analysis point estimate meets one of these equivalent endpoint-specific thresholds: (a) for mean change or rate of decline, 100 × (Wcontrol − Wtreatment) / Wcontrol ≥ 50, orienting both values so positive values mean worsening and requiring Wcontrol > 0; (b) for risk of a clinical-progression event, a risk ratio of at most 0.50 at the prespecified primary time point; or (c) for time to clinical progression, a hazard ratio of at most 0.50. Use the adjusted estimates if those are the prespecified primary estimates. An odds ratio alone, a percentage change in the final score rather than deterioration, or a percentage increase in time to progression is not interchangeable with these measures. Treatment-arm improvement with control-arm deterioration can qualify. The point estimate, not the confidence-interval bound, determines the 50% threshold; a separate statistical-significance threshold is not imposed.

Resolves YES when qualifying official publications establish both the approval and the qualifying pivotal-trial result. The percentage may be stated explicitly or calculated from the published primary-analysis estimates. Use the most precise official values available; do not round an estimate below 50% upward. If only an officially reported relative reduction of 50% or more is available, that report suffices unless more precise official figures show it falls below 50%.

Exclusions (resolve NO). Trial results without approval; investigational permissions, expanded access, emergency-use permissions, and regulatory submissions; secondary, exploratory, post-hoc, or open-label-extension analyses; and qualifying effects only in a dose arm or population outside the approved indication do not count. A later efficacy study of an already-approved treatment counts only if it supports a qualifying new approval or supplemental approval. Neither a manufacturer announcement nor a journal article can independently establish the approval or override the regulator's identification of the primary analysis.

Resolving source. FDA approval letters, approved prescribing information (especially Section 14, Clinical Studies), and multidisciplinary, clinical, or statistical reviews in Drugs@FDA; or European Commission authorisation decisions in the Union Register of medicinal products, together with EMA European Public Assessment Reports (EPARs), assessment reports, and Summary of Product Characteristics (especially Section 5.1). These official sources are narrower than Public confirmation (A2). Trial protocols, statistical analysis plans, and registry histories may corroborate prespecification but cannot replace the official efficacy and approval evidence. Apply the Resolution window and Resolution date (A3): resolve when the last necessary qualifying official publication is published. The approval must be newly documented within the resolution window; a republication of a pre-opening approval does not qualify. This is an approval event, not a published-statistic question under the Observation period convention; the supporting trial may have ended before opening. Subsequent withdrawal does not undo YES.