GLP-1 receptor agonists are a class of peptide-based medicines that mimic a gut hormone called glucagon-like peptide-1. Over the past several years, two members of this class, semaglutide and liraglutide, have been prescribed in growing numbers specifically for weight management. As prescriptions climb, researchers have been working hard to understand what happens to patients in real clinical settings, not just in tightly controlled trials.
A multicentre cohort study published in the journal Diabetes, Obesity and Metabolism examined electronic health records from 13 hospitals in South Korea, covering the period from 2018 to 2025. The researchers asked a straightforward question: compared with similar patients who did not start these peptides, what additional health risks, if any, did new users experience? The scale of the data gave the analysis unusual weight, with more than 75,000 matched individuals included.
Study design and why it matters
The research team used what is called a new-user design. That means they only counted patients from the moment they first started one of the two peptides, which avoids mixing in people who had already been on the medicine for months or years without incident. Electronic health records were standardised across all 13 hospitals using the OMOP Common Data Model, a widely used framework that allows different hospital systems to share data in a consistent format.
Patients who started semaglutide or liraglutide were paired with non-initiators through a statistical technique called propensity score matching. This attempts to create comparison groups that look similar on paper, balancing factors like age, existing conditions, and other medications, so that any differences in outcomes can be more fairly attributed to the peptide rather than to pre-existing differences between people who chose to take it and those who did not.
Ten specific safety outcomes were tracked. Each hospital calculated its own hazard ratios using Cox proportional hazards models, and those results were then pooled through meta-analysis. A hazard ratio above 1.0 means the outcome occurred more often in the peptide group than in the matched comparison group.
Psychiatric outcomes
One of the most notable clusters of findings involved mental health. For semaglutide initiators, the hazard ratio for psychiatric disorders overall was 2.02, meaning roughly twice the rate observed in non-initiators. Within that category, anxiety disorder carried a hazard ratio of 2.39 and depressive disorder reached 3.42, with confidence intervals that did not include 1.0, indicating statistical significance.
Liraglutide showed a similar but somewhat smaller pattern. The hazard ratio for psychiatric disorders overall was 1.66. Anxiety disorder came in at 1.68 and depressive disorder at 1.51. Both were statistically significant across the pooled analysis.
The researchers were careful not to claim that the peptides directly cause these conditions. Observational data can reveal associations, but confounding factors that propensity matching cannot fully remove may still be present. People seeking weight-loss treatment may already carry higher rates of mood-related conditions, or the stress of initiating a new medical regimen could play a role. What the study establishes is that the signal exists and is consistent enough to warrant attention.
Gastrointestinal and liver findings
GLP-1 receptor agonists are known to slow stomach emptying, which is part of how they reduce appetite. That same mechanism may explain why semaglutide was associated with a markedly elevated hazard ratio of 3.91 for gastrointestinal dysmotility or obstruction, conditions where the normal movement of food through the digestive tract is disrupted. The wide confidence interval (1.42 to 10.82) reflects genuine uncertainty, partly because this was a relatively rare outcome in absolute terms.
Liraglutide was linked to an increased risk of pancreatobiliary disorders, a category covering the pancreas, bile ducts, and gallbladder. The hazard ratio for this grouping was 1.33, and for the specific subset of pancreatitis, cholangitis, and cholelithiasis it was 1.40. Liraglutide initiators also showed an elevated hazard ratio of 1.46 for hepatic impairment, though the authors noted this finding was not significant in some sensitivity analyses, which introduces some caution.
These gastrointestinal signals are not entirely new to the literature. Earlier studies and regulatory labels have flagged gastrointestinal discomfort as a common side effect, and pancreatitis has been a monitored outcome since these peptide classes were first studied. What this cohort adds is a large, real-world confirmation using data from patients who were taking these medicines specifically for weight management rather than for blood sugar control.
Vision impairment signal
Both peptides were associated with elevated rates of vision impairment. For semaglutide the hazard ratio was 1.58, and for liraglutide it was 1.34. Both reached statistical significance in the primary analysis, though the semaglutide result was not consistent across all sensitivity analyses.
The biological reason for a possible link between GLP-1 receptor agonist use and eye-related outcomes is not fully established. Some researchers have speculated that rapid changes in blood sugar or blood pressure accompanying weight loss could affect the small blood vessels of the eye, particularly in people who already have some degree of underlying vascular vulnerability. The study did not explore mechanisms, so this remains an open question in the literature.
Heterogeneity between the two peptides
One of the more scientifically interesting aspects of the study is that semaglutide and liraglutide did not produce identical risk profiles. Both are GLP-1 receptor agonists and both are peptide-based, but they differ in their molecular structure, half-life, and dosing schedule. Semaglutide has a longer half-life and is typically dosed once weekly, while liraglutide requires daily dosing.
The gastrointestinal dysmotility signal was much stronger for semaglutide than for liraglutide. Conversely, hepatic impairment and pancreatobiliary signals appeared more clearly for liraglutide. This heterogeneity across agents suggests that grouping all GLP-1 receptor agonists together in safety discussions may obscure meaningful differences. The authors called for careful patient selection and monitoring, noting that the two medicines should not be assumed to carry identical risk profiles.
Limitations and context for the findings
The authors acknowledged several important limitations. Propensity score matching reduces confounding but cannot eliminate it entirely. Patients who seek out weight-loss prescriptions are a self-selected group, and their baseline characteristics, even after matching, may differ from non-initiators in ways the data cannot fully capture.
The study was conducted in South Korea, and it is not guaranteed that findings will translate identically to other populations with different genetic backgrounds, dietary patterns, or healthcare systems. The follow-up period varied across patients, and some of the outcome categories had relatively few events, which widens confidence intervals and introduces uncertainty.
Despite these caveats, the scale of the data and the consistency of results across sensitivity analyses make this one of the more robust observational assessments of GLP-1 receptor agonist safety outside of clinical trials. The literature now suggests that psychiatric outcomes, gastrointestinal complications, and vision changes are areas where ongoing surveillance is warranted for patients who initiate these peptides for weight management.
