I've been tracking EVOKE closely. A few important design points people should understand:

EVOKE enrolled patients with early AD (MCI due to AD or mild AD dementia) confirmed by amyloid PET or CSF biomarkers. This is critical — they're not studying general cognitive decline in obese patients; they're studying biomarker-confirmed Alzheimer's disease. The primary endpoint is change from baseline in CDR-SB (Clinical Dementia Rating Sum of Boxes) at 104 weeks.

What excites me is the secondary endpoints — they're measuring amyloid PET, tau PET, brain volume changes on MRI, plasma p-tau217, and NfL. So even if the clinical endpoint is ambiguous (CDR-SB is noisy), we'll have biomarker data to understand whether semaglutide is modifying disease biology.

On the observational data: I want to add the Husain et al. study from the UK Biobank that showed GLP-1 RA use was associated with reduced risk of dementia broadly (not just AD), with an adjusted OR of 0.53 (95% CI: 0.39-0.72).[4] The UK Biobank data is valuable because it has rich covariate adjustment including education, APOE status, and baseline cognitive scores.

sema had the largest effect size for AD risk reduction among the class

This could relate to CNS penetrance. Semaglutide's fatty acid side chain (C18 diacid) enables albumin binding that may facilitate transport across the blood-brain barrier via receptor-mediated transcytosis. Liraglutide (C16 chain) also crosses, but dulaglutide and exenatide LAR may have limited CNS exposure due to molecular size. The ELAD trial with liraglutide showed some signals on FDG-PET but was underpowered (n=204).[5]

I want to inject some healthy skepticism. The observational HRs of 0.47-0.53 are enormous effect sizes for a single drug class against a neurodegenerative disease. For context, anti-amyloid antibodies like lecanemab (Clarity AD trial) achieved a 27% slowing of clinical decline and that was considered groundbreaking.[6]

A 50%+ reduction in AD incidence would be the single greatest pharmacological advance in neurodegeneration history. That should make our priors skeptical.

Sources of potential confounding:

• Healthy user bias: GLP-1 RA patients may be more health-engaged, exercise more, have better diet quality
• Survival bias: GLP-1 RAs reduce CV mortality — patients surviving longer without stroke have more opportunity to receive an AD diagnosis, but the counterfactual matters
• Indication bias: Comparator groups often include sulfonylureas or insulin, which have their own metabolic risks (hypoglycemia is a known dementia risk factor)
• Detection bias: GLP-1 RA users may have more frequent medical contact, potentially leading to earlier diagnosis or different diagnostic patterns

The strongest counter-argument is the dose-response and duration-response relationships seen in some of these studies. Wang et al. found greater risk reduction with longer GLP-1 RA exposure (>2 years vs <1 year), which is harder to explain by confounding alone.

I'll withhold judgment until EVOKE reports. But I'm cautiously optimistic.

The "type 3 diabetes" angle is what convinced me this is biologically plausible. Suzanne de la Monte's group proposed the concept over a decade ago — AD brains show impaired insulin signaling, reduced glucose metabolism, and insulin resistance localized to hippocampal and cortical neurons.[7]

GLP-1 agonists normalize CNS insulin signaling through PI3K/Akt pathway activation, which in turn inhibits GSK-3beta (a kinase responsible for tau phosphorylation) and promotes neuronal survival via anti-apoptotic signaling. The mechanistic chain is actually quite well-worked-out in preclinical models.

Two mouse studies I find compelling:

• Liraglutide reduced cortical amyloid plaque burden by ~50% and improved spatial memory in APP/PS1 transgenic mice[8]
• Semaglutide reduced neuroinflammatory markers (GFAP, Iba-1) and improved synaptic density in 3xTg-AD mice after 10 weeks of treatment

Obviously mice are not humans. But the convergence of epidemiological signals, mechanistic plausibility, and preclinical proof-of-concept is compelling enough to warrant the Phase 3 investment Novo is making.

This is mind-blowing if it pans out. Can someone explain in simpler terms — are we saying that a weight loss drug could prevent Alzheimer's? How does that work?

My grandmother had AD and I've been on semaglutide for 8 months for weight loss. Should I feel "hopeful" about the brain protection angle, or is it too early?