For the past couple of years, one of the loudest concerns in the GLP-1 conversation has been neuropsychiatric safety. Headlines about depression, anxiety and suicidal thoughts dogged the class through 2025, and regulators kept the question on their books with formal reviews.
A new analysis published September 1 in npj Metabolic Health and Disease, a Nature Portfolio journal, cuts against that narrative in a specific and interesting way. Among 63,215 patients with pre-existing neuropsychiatric conditions, higher attained semaglutide doses were associated with lower incidence of most neuropsychiatric events over the following two years, independent of weight loss.
It is a counter-narrative, and it deserves to be looked at honestly, which means noting both what it found and what it did not.
The study design
The authors - Murugadoss, Venkatakrishnan and Soundararajan - ran an observational study using a large electronic health record-derived cohort. They identified 63,215 patients with pre-existing neuropsychiatric conditions who started semaglutide and then tracked 24 incident neuropsychiatric outcomes after treatment initiation.
The analysis worked in two phases. In a propensity-matched comparison, semaglutide was associated with broadly lower neuropsychiatric event risk over two years compared with three other diabetes drug classes: metformin, SGLT2 inhibitors and DPP-4 inhibitors. That part aligns with several prior observational studies, including one in JAMA Network Open in 2024 that found no increased risk of suicidality with GLP-1s and some signals of reduced risk of depression and anxiety.
The novel part was the dose analysis. Within the semaglutide-treated cohort, the researchers measured the highest dose each patient attained during the first two years after starting (the “pre-landmark period”), then looked at neuropsychiatric event incidence during the subsequent two years (the “post-landmark period”). Higher attained dose was associated with significantly lower incidence of:
- Substance-related disorders
- Mood disorders
- Anxiety and stress-related disorders
- CNS atrophies
- Neuromuscular disorders
- Eating, sleep and behavioural disorders
- Personality and impulse-control disorders
Seven of the diagnostic categories showed significantly lower incidence with higher dose. Notably, dementia and other CNS degenerative diseases showed no difference between high and low dose groups, consistent with prior trial data. And one detail stood out as genuinely thought-provoking: for most neuropsychiatric diagnoses, post-landmark incidence was strongly associated with the maximum attained dose, but incident cognitive symptoms and speech and language symptoms were more closely associated with weight loss.
That last point matters. It suggests different neuropsychiatric outcomes may track different mechanisms. Dose-dependent reductions in mood and anxiety events look like a drug effect. Cognitive symptom incidence tracking weight loss looks like it belongs to the metabolic side of the equation, which makes sense given how much of the semaglutide cognitive literature is tied to metabolic and vascular changes.
What the Research Says
- Observational cohort of 63,215 patients with pre-existing neuropsychiatric conditions, tracking 24 incident outcomes after semaglutide initiation.
- Propensity-matched analysis: semaglutide associated with lower neuropsychiatric event risk over two years versus metformin, SGLT2 inhibitors and DPP-4 inhibitors.
- Within the semaglutide cohort, higher attained dose in the first two years predicted significantly lower two-year incidence of substance-related disorders, mood disorders, anxiety and stress-related disorders, CNS atrophies, neuromuscular disorders, eating/sleep/behavioural disorders and personality/impulse-control disorders.
- Dementia and CNS degenerative disease incidence was similar between high and low dose groups (P = 0.15).
- Cognitive and speech/language symptoms tracked weight loss more closely than dose.
Why the dose finding is interesting
Observational data showing GLP-1 patients have lower risk of depression and anxiety than patients on other diabetes drugs is common in the literature. That pattern has held across several large cohorts by 2026. The dose gradient is the less common finding, because it is harder to explain away.
If semaglutide carried a neuropsychiatric hazard, you would generally expect the signal to get worse as the dose went up. Finding the opposite - lower incidence of mood, anxiety and substance-related disorders at higher doses - is the inverse of the hazard pattern. The obvious alternative explanation is selection: maybe the patients who tolerate higher doses are a healthier subset, or the sickest patients are down-titrated early. The propensity matching and landmark design attempt to blunt that, but they cannot eliminate it entirely in observational data.
There is also a plausible mechanism. GLP-1 receptor agonists modulate inflammation and dopaminergic signalling, both of which are implicated in mood disorders. Semaglutide has documented anti-inflammatory effects in animal models, including on the brain. If some of the mood benefit flows through inflammation reduction, then a dose-response relationship is biologically coherent rather than surprising.
The caveats, stated plainly
This is an observational study, and the authors’ own framing is appropriately careful. Electronic health record cohorts carry coding limitations, unmeasured confounding is always possible, and the landscape continues to shift as real-world usage expands.
Two specific limitations are worth highlighting. First, the population studied had pre-existing neuropsychiatric conditions, which is a strength for statistical power but means the findings may not generalise to people without such histories. Second, the analysis relies on prescribed dose as recorded in records, which may not perfectly reflect what patients actually took. Adherence in the real world is messier than in trials.
None of this means the study proves semaglutide is protective against psychiatric events. What it does is add a serious, well-powered data point to a question that has been dominated by fear-based framing for too long. The causal question can only be settled by randomised trials with neuropsychiatric endpoints, and there is an argument that the class deserves them given how widespread use has become.
Why this matters for the community
The peptide and GLP-1 research community has spent the better part of two years defending a nuanced position on neuropsychiatric safety: the signal in pharmacovigilance reports exists but has not translated into consistent causal evidence in controlled studies. This paper gives that position a concrete, citable dose-response observation.
It is also a useful case study in how to read a counter-narrative. The finding does not whitewash the class, the caveats are substantial, and the dementia result shows the analysis did not find benefits everywhere. That mix - real signal, plausible mechanism, honest limitations - is what good research communication looks like.
For Australian readers specifically, the regulatory landscape is relevant context. The TGA has issued safety communications about GLP-1s and psychiatric events, and the class carries ongoing monitoring. Findings like this are part of the evidence base regulators will weigh as those reviews continue to develop. None of this should be read as medical guidance, and as always the research-use distinction holds: grey area interest is in the science, not in dosing recommendations.
The semaglutide research page covers the compound in more depth, and the Telegram community is where new papers like this get discussed as they land.
Sources
Source: Murugadoss K, Venkatakrishnan AJ, Soundararajan V. Higher semaglutide dose is associated with lower neuropsychiatric event incidence independent of weight loss. npj Metabolic Health and Disease, 2026. DOI: 10.1038/s44324-026-00131-3
Disclaimer: This article is for educational and informational purposes only. It does not constitute medical advice, therapeutic recommendations, or endorsements of any compound. Grey Highway is a research-education community. We do not sell, supply, or promote the use of research compounds. Always consult a qualified healthcare professional regarding health decisions. For Australian regulatory information, visit the TGA website.