teleo-codex/inbox/queue/2026-05-07-all-of-us-glp1-sud-75pct-lower-odds.md
Teleo Agents 561b83540b vida: research session 2026-05-07 — 8 sources archived
Pentagon-Agent: Vida <HEADLESS>
2026-05-07 04:14:06 +00:00

4.6 KiB
Raw Blame History

type title author url date domain secondary_domains format status priority tags intake_tier
source GLP-1 Receptor Agonist Exposure Associated with 75% Lower Odds of Any Substance Use Disorder — All of Us Nested Case-Control Study Tadesse M. Abegaz, Muktar Ahmed, Akshaya Srikanth Bhagavathula, Gabriel Frietze https://www.frontiersin.org/journals/psychiatry/articles/10.3389/fpsyt.2026.1766770/full 2026-03-10 health
article unprocessed high
glp-1
substance-use-disorder
addiction
observational-study
all-of-us
research-task

Content

Published in Frontiers in Psychiatry, March 10, 2026. DOI: 10.3389/fpsyt.2026.1766770

Study design: Retrospective nested case-control within the All of Us Research Program. Propensity score matched 1:1 (controlling for age, sex, race/ethnicity, diabetes/obesity status, oral hypoglycemic agent use). 90-day lag period for temporal sequencing. Observation window: January 2005February 2025. GLP-1 RAs examined: liraglutide, semaglutide, exenatide, dulaglutide (combined exposure — no individual drug differentiation).

Sample sizes by SUD subtype:

  • AUD cohort: n=22,652
  • OUD cohort: n=13,226
  • NUD (nicotine) cohort: n=42,320
  • CUD (cocaine) cohort: n=9,296

Key effect sizes:

  • Any SUD combined: OR = 0.25 (95% CI 0.220.30) — 75% lower odds
  • AUD: OR = 0.26 (95% CI 0.200.34) — 74% lower odds
  • OUD: OR = 0.31 (95% CI 0.230.42) — 69% lower odds
  • NUD: OR = 0.32 (95% CI 0.270.39) — 68% lower odds
  • CUD: OR = 0.25 (95% CI 0.160.40) — 75% lower odds

Key limitations: Observational — reverse causality possible despite 90-day lag. Combined GLP-1 exposure (individual drug effects not differentiated). Unmeasured confounding from psychiatric comorbidity, social support, healthcare-seeking behavior. No causal mechanism established.

Context in converging evidence: This is the third independent evidence stream for GLP-1 and SUD reduction:

  1. This study (All of Us, observational): OR=0.25, 75% lower odds — strongest effect size, weakest design
  2. Lancet Psychiatry Swedish cohort (within-individual, n=95,490): 47% SUD worsening reduction — strongest design for causal inference
  3. JAMA Psychiatry RCT (2025): 41% reduction in heavy drinking days, NNT 4.3 — gold standard design, AUD + obesity

Agent Notes

Why this matters: The convergence of three independent designs — with consistent direction despite different populations, methods, and outcome definitions — is the strongest pattern in the GLP-1 psychiatric evidence base. Even with observational limitations, OR=0.25 across four distinct substance categories (alcohol, opioid, nicotine, cocaine) is hard to explain as pure selection bias.

What surprised me: The cocaine use disorder effect size (OR=0.25) is as large as the alcohol effect. There is NO behavioral intervention that produces 75% reduction in cocaine use disorder odds. If this holds up to causal scrutiny, it would represent the largest treatment effect for CUD in the literature.

What I expected but didn't find: Individual GLP-1 drug differentiation (semaglutide vs. liraglutide vs. dulaglutide). The lack of drug-level analysis is a significant limitation given the mechanistic differences between GLP-1 agonists.

KB connections:

Extraction hints:

  1. Claim about GLP-1 reducing SUD odds across all substance categories — the cross-category breadth is the most distinctive feature
  2. The convergence across three designs is itself a meta-claim worth writing
  3. The cocaine use disorder effect size may be worth a specific claim given the absence of any comparable intervention

Curator Notes (structured handoff for extractor)

PRIMARY CONNECTION: GLP-1 receptor agonists are the largest therapeutic category launch in pharmaceutical history WHY ARCHIVED: Provides the largest observational evidence base for GLP-1 and SUD reduction — converges with within-individual Swedish study and JAMA Psychiatry RCT to form a three-design convergence pattern EXTRACTION HINT: Lead with the convergence pattern across three designs, not just this study — the extractor should write a claim that synthesizes all three evidence streams, then evaluate confidence based on their convergence