GLP-1 psychiatric effects are directionally opposite in metabolic versus psychiatric disease patients — protective in metabolic cohorts but potentially harmful in severe psychiatric comorbidity with concurrent psychotropic use

The GLP-1 psychiatric safety paradox resolves through population stratification rather than dismissing either signal. Clinical trials and cohort studies systematically exclude patients with 'psychiatric instability' — specifically those with substance use disorders, prior mood episodes, or active anhedonia. This creates a bifurcated evidence base: (1) Trial/cohort populations over-represent metabolically driven psychiatric patients where GLP-1 appears protective (Swedish cohort showing reduced depression/anxiety in metabolic disease context), and (2) Pharmacovigilance captures real-world deployment including psychiatric comorbidity patients where GLP-1 may worsen symptoms. The highest-risk subpopulation is patients on concurrent psychotropic medications (antidepressants, benzodiazepines) showing OR 4.07-4.45 for suicidality reporting. The Novo Nordisk semaglutide MDD program (interim data late 2026) will provide the first prospective RCT evidence in psychiatric patients rather than metabolic patients with psychiatric comorbidities, serving as the decisive test of whether GLP-1 is genuinely antidepressant or whether the metabolic patient finding is a selection effect. The eating disorder signal is consistent with this framework: GLP-1 appetite suppression may trigger pathology in vulnerable patients systematically excluded from trials but present in real-world deployment.

Challenging Evidence

Source: WHO guideline 2025-12-01, absence of psychiatric contraindications

WHO guideline excludes only pregnant women as explicit contraindication, with no mention of psychiatric comorbidity screening despite documented eating disorder signal (aROR 4.17-6.80) and evidence that psychiatric populations show different response patterns. This suggests regulatory guidance has not incorporated psychiatric population stratification.

Supporting Evidence

Source: VigiBase temporal analysis, Clinical Nutrition 2025

Sensitivity analysis of 2.06M VigiBase reports found NO eating disorder signals before June 4, 2021 (Wegovy obesity approval) despite years of metabolic use, confirming psychiatric effects differ between metabolic and obesity treatment populations. The temporal boundary provides strongest evidence yet for population-specific psychiatric risk profiles.

Extending Evidence

Source: Gill et al., JAMA Psychiatry 2026

First RCT evidence that therapeutic doses in MDD population reduce motivation deficit (opposite of anhedonia induction). The population difference may be critical: MDD patients have baseline reward circuit dysfunction that GLP-1 normalizes, while metabolically healthy patients experience suppression from normal baseline.

Supporting Evidence

Source: Sa et al. (2026)

Meta-analyses show 'modest antidepressant effects, greater in type 2 diabetes populations' while observational data in obesity populations show '195% increased depression risk and 106% increased suicidal behavior risk.' This confirms directionally opposite effects by population, though confounding by indication complicates interpretation.

Supporting Evidence

Source: PMC systematic review + JAMA Psychiatry RCT

AUD RCT (N=108) showed 41.1% reduction in heavy drinking days with no psychiatric adverse events in comorbid AUD + obesity population. However, community-based cohort study of general GLP-1 prescription recipients found 195% increased MDD risk. This supports the claim that GLP-1 psychiatric effects differ by population: beneficial in addiction/metabolic comorbidity, potentially harmful in general metabolic-only populations. The literature is internally inconsistent, with systematic reviews finding both 'promising results for depression' and the 195% MDD risk signal.