GLP-1 Psychiatric Safety in Adolescents

A September 2026 retrospective cohort study examined psychiatric safety in adolescents with overweight or obesity who started semaglutide and liraglutide, plus tirzepatide, compared with adolescents receiving lifestyle intervention. After propensity-score matching, each group contained 9,222 people. The study reported lower recorded risks of suicide-related events and anxiety in the incretin-treatment group. Suicidal ideation was also lower, while the paper found no statistically significant difference for suicide attempt, insomnia, depression, or eating disorders^[1].

The result is reassuring within the limits of observational research, but it does not prove that these medicines reduce psychiatric risk. Electronic-health-record studies can balance measured differences between groups, yet they cannot fully account for why one adolescent received a medicine and another received lifestyle care. The authors also report that an active comparison with metformin did not show a significant difference.

Psychiatric safety of GLP-1 medicines in adolescents

The paper used deidentified records from the TriNetX Global Collaborative Network. It compared new users of semaglutide, liraglutide, or tirzepatide with adolescents receiving lifestyle intervention, then followed psychiatric outcomes from 30 days through 1,095 days after the index date^[1]. The primary outcome combined recorded suicidal ideation and suicide attempt. That definition matters because a combined outcome can move even when one of its components does not.

In the matched analysis, the hazard ratio for suicide-related events was 0.74, with a 95% confidence interval of 0.57 to 0.95. Recorded suicidal ideation had a hazard ratio of 0.73. Anxiety had a hazard ratio of 0.92. The paper found no significant group difference for suicide attempt, insomnia, depression, or eating disorders^[1]. These estimates describe associations in the database, not a direct drug effect.

What this study can and cannot answer

This was not a randomised trial. Propensity-score matching makes the comparison more alike on recorded baseline characteristics, but it cannot remove unrecorded factors such as care access, family support, the reason a clinician selected a treatment, or changes in symptoms that never entered the record. A lower recorded event rate may reflect some of those differences as well as treatment exposure.

The outcome window was broad, and the study used coded health-record events rather than structured psychiatric interviews. It therefore cannot show whether a medicine caused, prevented, or had no effect on an individual experience. The study also grouped several medicines together. Readers looking for medicine-specific background can compare the site entries for semaglutide and tirzepatide, but this paper does not provide a separate medicine-specific association for each one^[1].

How the result compares with earlier adolescent data

A 2024 JAMA Pediatrics retrospective study used the same broad network type and compared adolescents with obesity prescribed a GLP-1 receptor agonist with matched adolescents receiving lifestyle intervention. It reported suicidal ideation or attempts in 1.45% of the GLP-1 group and 2.26% of the comparison group over 12 months, corresponding to a hazard ratio of 0.67. Gastrointestinal symptom codes were more common in the GLP-1 group^[2].

The 2026 paper extends that signal by including dual GIP/GLP-1 treatment and by examining several psychiatric outcomes. Both studies remain observational comparisons against lifestyle care. Their agreement is useful context, but it does not replace a randomised design or establish a protective psychiatric effect. The newer paper's metformin comparison is a further reason to avoid reading the lifestyle comparison as a settled answer^[1][2].

What the paper adds to the safety discussion

Concerns about suicidality have made psychiatric outcomes a visible part of GLP-1 safety discussions. This study does not identify an increased psychiatric risk in its matched adolescent cohort. It also reports lower recorded rates for selected outcomes, with differences by sex and type 2 diabetes status in subgroup analyses^[1]. Subgroup findings can help frame future research, although they are less definitive than a primary analysis and may be affected by smaller event counts.

The most accurate reading is narrow: in this database study, incretin-based treatment was not associated with an increased rate of the psychiatric outcomes examined. The paper does not settle long-term psychiatric safety for every adolescent, compare individual medicines head to head, or establish that a medicine lowers suicide-related risk. Its results add evidence to an area that still needs careful follow-up.

The record-based design has another practical boundary. A coded event depends on contact with a health system and on how that contact was documented. A database can identify patterns across large groups, yet it cannot describe the context of every event or every change in mood. The study therefore belongs alongside, rather than above, trial safety data, regulatory monitoring, and ongoing research that separates individual medicines and uses comparators selected for the clinical question.

The authors describe the findings cautiously. Their sensitivity analyses supported the main result, but the discussion notes selection bias and the absence of a significant difference in the metformin comparison. That combination of a reassuring primary association and a nonconfirming active comparison is the central interpretive point. It supports continued scrutiny of the question rather than a broad claim about psychiatric benefit.

Medical disclaimer: This article is for educational and journalistic purposes only and does not constitute medical advice. PeptideMethods does not sell, distribute, or facilitate the sale of any product.