GLP-1s and Alzheimer's Disease: What the evoke Trials Actually Found

The evoke trials tested whether a GLP-1 medication could slow down Alzheimer's disease in people who already had early symptoms. Novo Nordisk ran two nearly identical phase 3 studies, evoke and evoke+, enrolling adults aged 55 to 85 who had either mild cognitive impairment (memory and thinking problems that do not yet interfere much with daily life) or mild dementia due to Alzheimer's. Every participant had to have confirmed amyloid in the brain, so this was a biologically verified Alzheimer's population, not a general group of forgetful older adults.

Participants were randomly assigned to take oral semaglutide 14 mg once daily, or a matching placebo pill, for two years. Oral semaglutide is the same molecule as the injections sold as Ozempic and Wegovy, delivered as a tablet. The two trials together enrolled more than 3,800 people, which makes this one of the largest tests of a repurposed metabolic drug in Alzheimer's ever run.

The primary endpoint was the change in CDR-SB score at week 104. CDR-SB stands for Clinical Dementia Rating - Sum of Boxes, an 18-point scale where a clinician scores memory, orientation, judgment, community affairs, home life and personal care. Higher scores mean more impairment, and a change of roughly 1 to 2 points over two years is typical in early Alzheimer's. The trials also tracked daily function using the ADCS-ADL-MCI scale, plus a panel of Alzheimer's blood biomarkers such as phosphorylated tau. If you want a refresher on how these drugs work in the body, our explainer on [how GLP-1 medications work](/blog/how-glp1-medications-work-mechanism-explained) covers the mechanism.

There was no difference between semaglutide and placebo on the main outcome. Over 104 weeks, mean CDR-SB scores worsened by 2.3 points on semaglutide versus 2.3 on placebo in one trial, and 2.2 versus 2.1 in the other. Those gaps are statistically and clinically meaningless. Both groups declined at the same pace.

The secondary endpoint told the same story. Change in the ADCS-ADL-MCI score, which measures whether someone can still manage things like paying bills, preparing meals and keeping appointments, also showed no separation between the drug and placebo arms in either trial.

The one place semaglutide did move the needle was biomarkers. Blood markers associated with Alzheimer's pathology improved in the treatment group. That is scientifically interesting, because it suggests the drug reached the brain and did something measurable. It just did not change how people thought, remembered or functioned.

Safety was consistent with what is already known about semaglutide. Gastrointestinal side effects such as nausea and reduced appetite were more common on the drug, which matters more than usual in an older population where unintended weight loss and muscle loss carry real risk. Nothing in the safety data suggested the drug harmed cognition.

The results were presented at the AD/PD 2026 international conference and published in The Lancet. Coming after the earlier topline announcement, the full publication removed most of the remaining ambiguity: this was a clean, well-powered null result across two independent trials.

The hypothesis was reasonable, and it came from several directions at once. GLP-1 receptors are not only in the pancreas and gut. They are also present on neurons in the hippocampus and other brain regions involved in memory. In animal models of Alzheimer's, GLP-1 drugs reduced brain inflammation, improved insulin signaling in neurons and, in some experiments, reduced amyloid plaque burden.

There was human signal too. A secondary analysis of three large type 2 diabetes trials of semaglutide found fewer dementia diagnoses among people taking the drug than expected, though dementia was not what those trials were designed to measure. Large observational database studies pointed the same way. Researchers also noted that type 2 diabetes, obesity and insulin resistance are all established risk factors for dementia, so a drug that improves all three could plausibly protect the brain over time.

That logic produced a genuine wave of optimism. Alzheimer's has an unusually thin treatment pipeline, and the anti-amyloid antibodies approved in recent years slow decline only modestly while requiring infusions and brain MRI monitoring. A cheap daily pill that also improved weight, blood sugar and cardiovascular risk would have been transformative.

The trials were the test of that hypothesis, and the hypothesis did not hold, at least not in this population at this stage of disease. This is what well-run phase 3 trials are for. Encouraging animal data and observational associations are hypotheses, not conclusions, which is worth remembering the next time a headline promises that a familiar drug prevents dementia.

The most likely explanation is timing. By the time someone has amyloid-confirmed mild cognitive impairment or mild dementia, Alzheimer's has typically been developing silently in the brain for 15 to 20 years. Substantial numbers of neurons and synapses are already lost, and lost neurons do not come back. A drug that dampens inflammation or improves metabolic signaling may simply arrive too late to change the trajectory.

A second explanation is that the improved biomarkers were not the right biomarkers. Alzheimer's research has repeatedly learned that moving a marker is not the same as moving the disease. Several drugs have cleared amyloid impressively with little or no clinical benefit. Biomarkers are useful for showing that a drug is biologically active, but they are not a substitute for measuring whether people can still recognize their grandchildren or manage their own medications.

There is also a dose and delivery question. Oral semaglutide has substantially lower and more variable absorption than the injectable form, and it must be taken fasting with a small sip of water. Whether a higher dose, or an injectable GLP-1, or a dual agonist like tirzepatide would reach brain tissue differently is still unanswered.

Finally, Alzheimer's is probably not one disease. Researchers increasingly expect that combination therapy, targeting amyloid, tau, inflammation and metabolism together, will be needed, much as with cancer or HIV. Several commentators framed the evoke results not as the end of metabolic approaches but as evidence that a single agent used alone, late in the disease, is unlikely to be enough.

No. A null result means the drug did not help in this specific situation. It does not mean the drug caused harm. In evoke and evoke+, people taking semaglutide declined at the same rate as people taking placebo, not faster.

That distinction matters because a lot of people taking GLP-1 medications for weight or diabetes saw the headlines and worried they were doing something risky. The trial data does not support that worry. The known safety profile of semaglutide, including gastrointestinal effects, gallbladder risk and the rare eye complication described in our piece on [semaglutide and NAION vision risk](/blog/semaglutide-and-vision-naion-eye-risk-explained), is unchanged by these results.

It is also worth keeping the broader evidence in view. Semaglutide reduced major cardiovascular events by about 20% in the SELECT trial (NEJM, 2023), a finding covered in our summary of [what the SELECT trial found](/blog/glp1-heart-health-what-the-select-trial-found). Cardiovascular health and brain health are deeply linked over decades, because most dementia in older adults involves some vascular contribution alongside Alzheimer's pathology. A drug that lowers cardiovascular risk is plausibly doing something useful for long-term brain aging, even if it cannot reverse established Alzheimer's.

What these results do rule out is the specific hope that starting a GLP-1 after an Alzheimer's diagnosis will slow the disease. If you or a family member were considering a GLP-1 for that reason, this is the evidence to discuss with a neurologist.

It means the levers that matter most are still the ones you already have, not a pill. Roughly 45% of dementia cases worldwide are attributable to modifiable risk factors, according to the 2024 Lancet Commission on dementia prevention. The largest of those in midlife are hearing loss, high LDL cholesterol, high blood pressure, physical inactivity, obesity, smoking, excess alcohol, depression and social isolation.

For women specifically, midlife is when several of these shift at once. Blood pressure and LDL cholesterol both tend to rise across the menopause transition, as covered in [why LDL rises after 45](/blog/menopause-and-cholesterol-why-ldl-rises-after-45) and [why blood pressure climbs in menopause](/blog/menopause-blood-pressure-why-it-rises-and-what-helps). Treating those is real dementia risk reduction, with far better evidence than any single drug currently has.

The brain fog many women experience in perimenopause is a separate issue and is usually not an early sign of dementia. It typically improves after the transition. Our guide to [menopause brain fog](/blog/menopause-brain-fog-why-it-happens-and-what-helps) walks through what helps.

On hormone therapy and dementia, the evidence remains genuinely mixed and depends heavily on timing and formulation. That question is unpacked in [menopause and dementia risk](/blog/menopause-and-dementia-risk-does-estrogen-protect-your-brain). The honest summary is that HRT should be prescribed for symptoms and quality of life, not as a dementia prevention strategy.

If you are on a GLP-1 for weight or metabolic reasons, keep taking it for those reasons. Just do not count it as brain insurance.

Researchers are moving in three directions. The first is earlier intervention. Because the timing explanation is the most plausible one, the obvious next question is whether a metabolic drug given to cognitively normal people at high risk, years before symptoms, changes anything. Prevention trials like this are slow and expensive, because you have to follow thousands of healthy people for many years, but they are the only way to test the hypothesis fairly.

The second is combination therapy. Commentary from the Alzheimer's Drug Discovery Foundation following the readout emphasized that Alzheimer's likely needs multi-target treatment, pairing anti-amyloid or anti-tau agents with drugs that address inflammation and metabolic dysfunction. A GLP-1 could still have a role as part of a stack even if it fails alone.

The third is better patient selection. It is possible that a subgroup, such as people with Alzheimer's plus type 2 diabetes or significant insulin resistance, would respond differently. Subgroup analyses from evoke will be scrutinized closely, though such findings are hypothesis-generating only.

Meanwhile, the wider GLP-1 pipeline continues to expand into other conditions, from sleep apnea to fatty liver to osteoarthritis. Our roundups of [orforglipron, the first oral GLP-1 pill](/blog/orforglipron-foundayo-oral-glp1-pill-attain-results) and [what STEP UP found with semaglutide 7.2 mg](/blog/semaglutide-7-2-mg-step-up-trial-results-explained) track where the class is actually delivering.

For now, the accurate statement is simple: GLP-1s are excellent metabolic drugs and, on current evidence, not Alzheimer's drugs.