GLP-1s and the brain: Parkinson's, Alzheimer's and mood, between signal and trial

Why the brain

The hypothesis began in the lab: exenatide and lixisenatide showed neuroprotective effects in animal models of Parkinson's disease, and liraglutide in Alzheimer's models¹²⁴. Then came data from patients with type 2 diabetes, in whom GLP-1 users were associated with fewer dementia diagnoses⁶⁸. On that basis, randomized trials were launched in neurodegenerative diseases, something unusual for a metabolic drug.

Parkinson's: three trials, two directions

• Exenatide, Lancet 2017 (62 patients with moderate Parkinson's, 48 weeks plus 12 weeks of washout): at 60 weeks, the off-medication MDS-UPDRS III improved by 1.0 point with exenatide and worsened by 2.1 with placebo; adjusted difference −3.5 points (p = 0.0318). The authors noted that an effect on the disease could not be distinguished from a long-lasting symptomatic effect¹.
• Lixisenatide, NEJM 2024 (LIXIPARK) (156 patients diagnosed less than 3 years earlier, 12 months): MDS-UPDRS III changed by −0.04 with lixisenatide vs +3.04 with placebo; difference 3.08 (95% CI 0.86 to 5.30; P = 0.007). Other secondary end points did not differ substantially; nausea occurred in 46% and vomiting in 13%².
• Exenatide, Lancet 2025 (phase 3) (194 patients, 96 weeks, six UK hospitals): off-medication MDS-UPDRS III worsened by 5.7 points with exenatide and 4.5 with placebo; coefficient 0.92 (95% CI −1.56 to 3.39; p = 0.47). The authors concluded there is no evidence to support exenatide as a disease-modifying treatment for Parkinson's³. In June 2026, The Lancet published an expression of concern about this paper¹⁴.

The largest and longest trial was the negative one. Its own authors propose studying agents with better target engagement, or specific patient subgroups, to establish whether the class has any role in Parkinson's³. Lixisenatide remains an isolated positive signal, awaiting the "longer and larger trials" its investigators call for².

Alzheimer's: ELAD and EVOKE

ELAD (phase 2b, 204 patients without diabetes with mild to moderate Alzheimer's, 52 weeks of liraglutide) missed its primary outcome: cerebral glucose metabolism on PET did not differ between groups. A secondary executive scale (ADAS-Exec) favored liraglutide (0.15; 95% CI 0.03 to 0.28; unadjusted p = 0.01), with no difference in CDR-SB⁴.

EVOKE and EVOKE+ were the definitive test: two phase 3 trials at 566 sites in 40 countries, 3,808 patients aged 55 to 85 with amyloid-confirmed Alzheimer's at the mild cognitive impairment or mild dementia stage, given oral semaglutide up to 14 mg daily or placebo. At 104 weeks, CDR-SB worsened equally in both arms: difference −0.08 (p = 0.57) in EVOKE and 0.10 (p = 0.46) in EVOKE+. Both trials were discontinued due to negative clinical outcome. Safety was as expected for semaglutide⁵.

The observational signal is still there

In type 2 diabetes the data are consistent. In the pooled analysis of three double-blind cardiovascular trials, dementia was less frequent with GLP-1s than with placebo (HR 0.47; 95% CI 0.25 to 0.86), and in a Danish registry of 120,054 patients each additional year of exposure was associated with an HR of 0.89⁶. A meta-analysis of randomized trials published in JAMA Neurology found a reduction in dementia with GLP-1s (OR 0.55) that was not seen with SGLT2 inhibitors⁷.

Target-trial emulations point the same way. Using health records of 1,094,761 US patients with type 2 diabetes, semaglutide was associated with fewer first Alzheimer's diagnoses, with an HR of 0.33 vs insulin and 0.59 vs other GLP-1s⁸. In 2026, among 1,320 patients with type 2 diabetes and mild cognitive impairment, starting a GLP-1 rather than a DPP-4 inhibitor was associated with less dementia (HR 0.74) over up to 5 years, with a number needed to treat of 21⁹.

Why cohorts and trials disagree

The positive signal comes from people with type 2 diabetes, without dementia, treated for years. EVOKE studied a different population: already symptomatic, amyloid-confirmed Alzheimer's, and 86% without type 2 diabetes⁹. The authors of the 2026 study note that it was unclear whether GLP-1s influence progression to dementia in people with type 2 diabetes and mild cognitive impairment, and they describe their own work as hypothesis-generating⁹. The authors of the 2024 emulation also call for clinical trials to confirm the finding⁸.

Mood and suicidality: the alarm that was not confirmed

In July 2023 the EMA opened a review after reports of suicidal thoughts with liraglutide and semaglutide. In April 2024 its pharmacovigilance committee (PRAC) concluded that the available evidence does not support a causal association between GLP-1s (dulaglutide, exenatide, liraglutide, lixisenatide and semaglutide) and suicidal or self-injurious thoughts and actions¹⁰.

• In a cohort of 240,618 patients with overweight or obesity, semaglutide was associated with a lower risk of incident (HR 0.27) and recurrent (HR 0.44) suicidal ideation vs other anti-obesity drugs, with findings replicated in type 2 diabetes¹¹.
• In Swedish and Danish registers, with 124,517 GLP-1 users vs 174,036 SGLT2 users, the HR for suicide death was 1.25 (95% CI 0.83 to 1.88), not significant; for suicide or self-harm it was 0.83, and for incident depression or anxiety, 1.01¹².
• In January 2026 the FDA analyzed 91 placebo-controlled trials with 107,910 patients, found no increased risk and asked for the warning to be removed from the Saxenda, Wegovy and Zepbound labels¹³.

The effect on alcohol use, the other major central nervous system signal, has already been tested in randomized trials and is covered in a separate article on this blog.

What to tell the patient who asks

On current data, one can say that in people with type 2 diabetes GLP-1s were associated with less dementia in cardiovascular trials and large cohorts⁶⁷⁸; that they have not been shown to slow an already diagnosed Alzheimer's disease⁵; that in Parkinson's there is one positive trial and one negative phase 3²³; and that regulators found no evidence of an increased suicidality risk¹⁰¹³. It is a reassuring profile on safety and an open question on neurological efficacy.