---
type: source
title: "Psychiatric Adverse Events Linked to GLP-1 Analogues: A Disproportionality Analysis in American, Canadian and Australian Adverse Event Databases"
author: "PMC (multiple authors)"
url: https://pmc.ncbi.nlm.nih.gov/articles/PMC12630159/
date: 2025-01-01
domain: health
secondary_domains: []
format: paper
status: unprocessed
priority: medium
tags: [glp1, pharmacovigilance, eating-disorders, faers, disproportionality, psychiatric-adverse-events, cross-national]
intake_tier: research-task
---

## Content

Cross-national pharmacovigilance disproportionality analysis using three national adverse event databases: FAERS (US), Canada Vigilance Adverse Reaction Online Database (CVAROD), and Database of Adverse Event Notifications (DAEN, Australia).

Key findings on eating disorders:
- Dulaglutide in FAERS: ROR = 1.47 (95%CI 1.26-1.71)
- Dulaglutide in DAEN (Australia): ROR = 17.66 (95%CI 2.45-127.37) — extremely elevated
- Tirzepatide in FAERS: ROR = 1.58 (95%CI 1.14-2.20)
- Note: Lower values than VigiBase study — reflects different database populations, denominator calculations, and the fact that this is US/CA/AU only vs. global VigiBase

Indication bias explicitly acknowledged: "The databases used in this study did not contain information on any pre-existing psychiatric conditions in patients reporting AEs" — researchers could not "distinguish between a medicine-induced reaction and an event related to a patient's ongoing health issues"

Causality conclusion: Researchers "cannot determine a causal relationship between medication use and AEs" — results should be "interpreted with caution and supplemented with clinical studies to confirm associations"

## Agent Notes
**Why this matters:** Confirms the eating disorder signal across US, Canadian, and Australian databases — the signal isn't US-specific or FAERS-specific. The Australian DAEN result (ROR 17.66 for dulaglutide) is particularly high, suggesting the signal may be even stronger in populations where dulaglutide has higher relative prescribing share. The cross-national consistency increases biological plausibility.

**What surprised me:** The ROR values here (1.47-1.58) are MUCH lower than the VigiBase aROR (4.17-6.80). This discrepancy is methodologically significant: VigiBase uses adjusted analysis (aROR controls for co-reported adverse events) while this study may not. The VigiBase analysis is the more reliable estimate — but the discrepancy between studies is itself a finding that an extractor should note.

**What I expected but didn't find:** CVAROD (Canadian) results for eating disorders — the paper either didn't find a significant signal there or the abstract/summary I accessed didn't report it.

**KB connections:**
- [[AI diagnostic triage achieves 97 percent sensitivity]] — parallel: pharmacovigilance databases have the same "signal vs. noise" sensitivity-specificity tradeoff
- [[healthcare AI regulation needs blank-sheet redesign]] — pharmacovigilance methodology also needs redesign for medication classes with behavioral effects

**Extraction hints:** Useful as CORROBORATING evidence for the VigiBase signal across additional databases. The Australian DAEN outlier (ROR 17.66) deserves specific mention — possible explanation is small denominator, different population, or higher awareness/reporting in Australia. NOT sufficient as a standalone source for the eating disorder claim — use alongside VigiBase as supporting evidence.

**Context:** This is a methodologically weaker study than the VigiBase analysis but provides geographic breadth. The indication bias limitation is shared by all pharmacovigilance databases.

## Curator Notes (structured handoff for extractor)
PRIMARY CONNECTION: [[GLP-1 receptor agonists are the largest therapeutic category launch in pharmaceutical history but their chronic use model makes the net cost impact inflationary through 2035]]
WHY ARCHIVED: Corroborating cross-national evidence for the eating disorder pharmacovigilance signal — extractor should use as secondary evidence, not primary
EXTRACTION HINT: Flag the discrepancy between this study's ROR (1.47-1.58) and VigiBase's aROR (4.17-6.80) in any claims that cite both — the methodological difference matters.