Can GLP-1 Medicines Lower Dementia Risk? What the EVOKE Trials Changed

For about three years, GLP-1 medicines looked like they might do something no drug had managed: lower the odds of dementia as a side benefit of treating weight and blood sugar.

The early signals were striking. Health record studies reported Alzheimer's diagnoses falling by half or more among semaglutide users. A pooled look at three large heart trials found dementia diagnosed about half as often in the GLP-1 groups. Then, in November 2025, the first large randomized test arrived. Novo Nordisk's EVOKE and EVOKE+ trials gave oral semaglutide to 3,808 people with early Alzheimer's disease for two years, and the drug did not slow their decline.

Both of those things are true at once, and the gap between them is worth understanding. This article walks through each line of evidence, what it can and cannot show, which dementia risk factors GLP-1 medicines actually move, and why none of it says anything about microdose protocols.

Why did researchers think GLP-1 drugs might protect the brain?

Receptors in the brain, insulin, and inflammation

Researchers expected GLP-1 drugs to help the brain because GLP-1 receptors sit on neurons and immune cells in regions tied to memory, and because Alzheimer's disease shares features with metabolic disease. Brain tissue from people with Alzheimer's often shows poor insulin signaling, which some researchers have nicknamed type 3 diabetes. In mouse models, GLP-1 agonists reduced markers of brain inflammation, lowered amyloid plaque load, and protected synapses.

The vascular route made the case look stronger still. Type 2 diabetes is linked with roughly 1.5 to 2 times the dementia risk in large cohort studies, and small strokes and damaged blood vessels contribute to a large share of dementia in older adults. GLP-1 medicines lower the rate of major cardiovascular events, including stroke, in outcome trials such as SUSTAIN-6 and SELECT. A drug that protects arteries, lowers glucose, and calms inflammation looked like a sensible candidate.

What did the observational studies show?

Large associations from health records

Observational studies showed large associations between GLP-1 use and fewer dementia diagnoses. In 2024, Wang and colleagues at Case Western Reserve University analyzed US electronic health records of about 1 million people with type 2 diabetes and found semaglutide users had a 40 to 70 percent lower rate of a first Alzheimer's diagnosis over three years than users of insulin, metformin, and several other diabetes drugs (Alzheimer's & Dementia, 2024).

A Danish registry analysis published by Nørgaard and colleagues in 2022 pointed the same way. Dementia rates fell by about 11 percent for each additional year of GLP-1 exposure in people with type 2 diabetes. Those numbers come from comparing groups of real patients, which is both their strength and their weakness: the data are large and realistic, but the people who receive newer, pricier drugs differ from those who do not in ways no adjustment fully captures.

What did the pooled heart trials add?

Randomized patients, but dementia was not the question

The pooled heart trials added a randomized comparison, though dementia was never what those trials set out to measure. Nørgaard and colleagues combined LEADER (liraglutide), SUSTAIN-6 (injectable semaglutide), and PIONEER 6 (oral semaglutide), totalling 15,820 people with type 2 diabetes and high cardiovascular risk. Dementia was diagnosed in 15 people on a GLP-1 versus 32 on placebo, a hazard ratio of 0.47.

Forty-seven cases across nearly 16,000 people is a thin base. The trials did not test memory, so a dementia diagnosis depended on a doctor noticing and recording it as an adverse event. The randomization was real, however, which made this the strongest piece of supportive evidence before EVOKE and a key reason Novo Nordisk funded a dedicated phase 3 program.

What happened in the ELAD liraglutide trial?

Missed its main goal, with hints in the secondary results

The ELAD trial missed its primary endpoint but reported encouraging secondary results. Led by Paul Edison at Imperial College London, ELAD randomized 204 people with mild to moderate Alzheimer's disease to daily liraglutide or placebo for 52 weeks. The main measure, brain glucose metabolism on PET scans, did not differ between groups.

Secondary measures were more interesting. People on liraglutide showed about 18 percent less decline on a composite of memory and executive function, and MRI scans showed roughly half the loss of brain volume in temporal and frontal regions. Results presented at the 2024 Alzheimer's Association International Conference drew attention, but a phase 2 trial with a missed primary endpoint is a reason to run a bigger trial, not a conclusion on its own.

What did EVOKE and EVOKE+ find?

The largest test so far, and a clear negative

EVOKE and EVOKE+ found that oral semaglutide did not slow Alzheimer's progression. The twin phase 3 trials enrolled 3,808 adults aged 55 to 85 with mild cognitive impairment or mild dementia and confirmed amyloid pathology, and gave them oral semaglutide titrated to 14 mg daily or placebo for 104 weeks. The primary endpoint was the Clinical Dementia Rating Sum of Boxes (CDR-SB), the same scale used to judge the approved anti-amyloid antibodies.

In its November 24, 2025 announcement, Novo Nordisk reported no statistically significant difference on CDR-SB between semaglutide and placebo in either trial. Semaglutide did improve several Alzheimer's-related biomarkers, and it was generally well tolerated. The company ended the planned one-year extension of both trials.

Some researchers have pointed out that oral semaglutide at 14 mg produces lower blood levels than the injectable 2.4 mg weight-loss dose, and that brain penetration of semaglutide is limited. Those are fair questions for future trials. They do not change the plain result: in people who already had early Alzheimer's, the drug as tested did not slow the disease.

Why do the observational studies and EVOKE disagree?

Different people, different questions, different methods

The observational studies and EVOKE disagree mostly because they asked different questions of different people. The health record studies followed adults with type 2 diabetes, mostly without cognitive symptoms, and counted new diagnoses. EVOKE gave the drug to people whose Alzheimer's was already underway and measured whether decline slowed. Preventing a disease and slowing it once established are not the same task.

Bias explains part of the gap too. In health records, patients prescribed semaglutide tend to have better insurance, more regular follow-up, and fewer other illnesses than those prescribed insulin, and those advantages alone predict fewer dementia diagnoses. Some researchers also suspect reverse causation: people in the earliest, unrecognized stages of dementia may be less likely to start a new injectable drug. Randomized trials remove those effects, which is why a single large negative trial outweighs many positive associations.

Which dementia risk factors do GLP-1 medicines actually change?

Several, though none is the biggest lever

GLP-1 medicines change four of the 14 modifiable dementia risk factors named by the 2024 Lancet Commission: obesity, diabetes, hypertension, and high LDL cholesterol. In STEP 1, semaglutide 2.4 mg lowered systolic blood pressure by 6.2 mmHg and body weight by about 15 percent over 68 weeks. In people with type 2 diabetes, GLP-1 drugs typically lower HbA1c by 1 to 2 percentage points.

The Commission's estimates put those four factors together behind roughly 12 percent of dementia cases worldwide. Hearing loss and high LDL cholesterol alone each account for about 7 percent. Lowering any of these risks is good for general health. Whether lowering them with a GLP-1 translates into less dementia has not been tested, and the 45 percent figure applies to all 14 factors combined, not to anything a single drug can reach.

What actually has evidence for lowering dementia risk?

The Lancet Commission's highest-yield targets

The best supported ways to lower dementia risk target the factors the Lancet Commission ranks highest, and most of them involve no medicine at all. The list below follows the Commission's 2024 recommendations. None has the certainty of a drug approval, but each rests on decades of cohort data and, for blood pressure, randomized evidence from SPRINT MIND.

Does a microdose GLP-1 protocol protect the brain?

There is no evidence either way

There is no evidence that a microdose GLP-1 protocol protects the brain or lowers dementia risk. Every study discussed here used standard doses, and the only large trial built to answer the question used a full oral dose and found no slowing of Alzheimer's. If a full dose did not change the course of early Alzheimer's, there is no basis for expecting a smaller one to.

What lower-dose approaches can reasonably aim at is the metabolic side: steadier weight, blood pressure, and glucose with fewer gastrointestinal side effects. Those are worthwhile goals on their own terms. Marketing that links microdosing to memory, focus, or dementia prevention goes beyond anything the research supports, and anyone noticing memory changes should get them assessed rather than self-treat.

Is research on GLP-1 drugs and dementia over?

Narrower, not finished

Research on GLP-1 drugs and dementia is narrower after EVOKE but not finished. The open questions now center on prevention in people without symptoms, on drugs that reach the brain more easily, and on subgroups such as people with type 2 diabetes, where vascular damage plays a larger role. Dual and triple agonists like tirzepatide and retatrutide have not been tested for cognition in dedicated trials.

The lesson from the past three years is about evidence more than about any one drug. Large associations in health records generated a hypothesis, a small trial produced mixed signals, and a large randomized trial gave a clear answer to the question it asked. Anyone reading future headlines about GLP-1 medicines and the brain can ask which of those three kinds of study the headline is based on.

Frequently Asked Questions