22 findings across 8 drug records. Domain pages are descriptive indexes, not advice.
Most represented records
GLP-1 RA class (7)
Semaglutide (6)
Exenatide (4)
Tirzepatide|semaglutide (head-to-head) (1)
Liraglutide (1)
Evidence spread
High evidence 3Moderate evidence 7Low evidence 9Very low evidence 3
Tirzepatide|semaglutide (head-to-head)MARKETED
cns-neuropsychiatricMixed
Low evidence
ADDICTION (alcohol) - a within-class gradient: in a target-trial-emulation EHR study, tirzepatide and semaglutide (but NOT liraglutide or dulaglutide) were associated...
SourceHenney AE, Riley DR, Heague M et al., Diabetes Obes Metab, 2025Source
Full findingADDICTION (alcohol) - a within-class gradient: in a target-trial-emulation EHR study, tirzepatide and semaglutide (but NOT liraglutide or dulaglutide) were associated with lower incident AUD vs DPP-4 inhibitors, suggesting potency or GIP co-agonism may matter (an open question, not a verdict).
PopulationHenney target-trial-emulation observational cohort (US EHR >120M): T2D without prior AUD, propensity-matched vs DPP4i; tirz n=7165, sema n=20198, lira n=6565, dula n=19061; 18 months.
Fundingacademic / investigator-initiated (Henney et al., University of Liverpool; EHR target-trial emulation)
ADDICTION (opioid/cross-substance): in large EHR cohorts of patients with opioid or alcohol use disorder, GIP/GLP-1 receptor agonist prescriptions were associated with...
SourceQeadan F, McCunn A, Tingey B, Addiction, 2025 (+ exploratory OUD+AUD cohort, Commun Med 2...Source
Full findingADDICTION (opioid/cross-substance): in large EHR cohorts of patients with opioid or alcohol use disorder, GIP/GLP-1 receptor agonist prescriptions were associated with lower rates of opioid overdose and alcohol intoxication. The opioid evidence is OBSERVATIONAL-ONLY (no OUD RCT exists).
PopulationQeadan EHR cohorts (Oracle/Cerner, >100M): N=503,747 OUD + 817,309 AUD (2014-2022); plus an exploratory N=107,217 OUD+AUD-comorbid cohort.
Fundingacademic/independent analysis (manufacturer not study sponsor; per-study COI not individually audited)
Scope limitsfew events/wide CI; possible confounding by indication; surrogate/exploratory endpoint
Comparatorsno GIP/GLP-1 RA prescription
GLP-1 RA classMARKETED
cns-neuropsychiatricDecrease
Very low evidence
ADDICTION MECHANISM (RODENT, not human): GLP-1RAs act on the mesolimbic reward circuit - semaglutide reaches the nucleus accumbens and reduces alcohol-evoked dopamine...
SourceAranas C et al., EBioMedicine, 2023 (semaglutide/alcohol/NAcc dopamine); Herman RJ et al....Source
Full findingADDICTION MECHANISM (RODENT, not human): GLP-1RAs act on the mesolimbic reward circuit - semaglutide reaches the nucleus accumbens and reduces alcohol-evoked dopamine release (the canonical reward signal) and intake/relapse; liraglutide reduces nicotine self-administration and relapse-like reinstatement. A biologically-plausible substrate (in rodents) for the human craving/consumption signals - plausibility only, NOT shown to be the operative human mechanism, and it does NOT raise the grade of the human observational rows it is cross-referenced from.
PopulationRats and mice (Aranas semaglutide-alcohol; Herman liraglutide-nicotine); self-administration / microdialysis / locomotion models.
Fundingacademic (Aranas/Herman rodent neuroscience groups; no sponsor disclosure extracted)
Scope limitsanimal data; human relevance uncertain
Comparatorsvehicle/saline
GLP-1 RA classMARKETED
cns-neuropsychiatricDecrease
Low evidence
NEURODEGENERATION - the EPIDEMIOLOGY pole (in tension with the negative RCTs): in a Scandinavian active-comparator cohort, GLP-1RA use was associated with lower...
SourceEngstrom A, Svanstrom H, Hviid A et al., Diabetes Obes Metab, 2026 (Scandinavian PD cohort)Source
Full findingNEURODEGENERATION - the EPIDEMIOLOGY pole (in tension with the negative RCTs): in a Scandinavian active-comparator cohort, GLP-1RA use was associated with lower incident Parkinson's than sulfonylurea use. This is a lower-INCIDENCE association in diabetics (exposure 73% liraglutide), NOT disease modification - the randomised tests (exenatide PD ph3, EVOKE) were NEGATIVE, and reverse causation (prodromal disease reducing drug initiation) is a MATERIAL threat. The mechanistic rationale (brain GLP-1 receptors; reduced amyloid/tau, neuroinflammation; improved central insulin signalling in rodent models) is preclinical and did NOT translate to disease modification in the human RCTs.
PopulationEngstrom Scandinavian cohort (DK/NO/SE registers): N=158,961 new GLP-1RA vs 188,065 new sulfonylurea users, age >=45; exposure dominated by liraglutide (73%).
Fundingacademic/independent analysis (manufacturer not study sponsor; per-study COI not individually audited)
Scope limitsanimal data; human relevance uncertain; possible confounding by indication
Comparatorssulfonylureas
GLP-1 RA classMARKETED
cns-neuropsychiatricNo change
Very low evidence
PSYCHIATRIC SAFETY - consolidated verdict (cross-refs the existing safety-other MH rows, not re-gathered): the suicidality/self-harm signal was INVESTIGATED by...
SourceConsolidation of existing safety-other MH rows; new RCT-pooled suicidality meta-analysis ...Source
Full findingPSYCHIATRIC SAFETY - consolidated verdict (cross-refs the existing safety-other MH rows, not re-gathered): the suicidality/self-harm signal was INVESTIGATED by regulators (EMA PRAC, FDA) and LARGELY CLEARED in RCT and meta-analytic data (FDA removed the suicidal-ideation warning from liraglutide/semaglutide-2.4/tirzepatide labels), with residual pharmacovigilance (FAERS/VigiBase) noise and confounding-by-indication in observational data; RCT depression/anxiety symptoms did not increase.
PopulationAdults with T2D/obesity across the class; regulatory reviews, the STEP/SURMOUNT RCT programmes, RCT + observational meta-analyses (27-RCT pool; 82-study synthesis), pharmacovigilance databases.
Fundingacademic/independent analysis (manufacturer not study sponsor; per-study COI not individually audited)
Scope limitspossible confounding by indication
Comparatorsplacebo; active comparators; non-GLP-1 anti-obesity medications
GLP-1 RA classMARKETED
cns-neuropsychiatricNot directional
Low evidence
FOLK-CLAIM VERDICT ('one drink hits like three' / wrecked alcohol tolerance, described as near-universal for retatrutide): NOT SUPPORTED, likely FALSE as stated - and...
SourceQuddos F et al. GLP-1 receptor agonists and acute alcohol pharmacokinetics/subjective eff...Source
Full findingFOLK-CLAIM VERDICT ('one drink hits like three' / wrecked alcohol tolerance, described as near-universal for retatrutide): NOT SUPPORTED, likely FALSE as stated - and DISTINCT from the real, separately-evidenced reduced-craving effect. The only human alcohol-challenge data show LOWER blood-alcohol and LESS intoxication per drink, and the gastric-emptying physiology points the same way. The lived experience is better explained by drinking on a near-empty stomach, far lower desire (stopping at one), GI upset making a drink unpleasant fast, and (chronically) a smaller body-water volume from weight loss - none of which is 'higher blood alcohol per drink'. NO retatrutide-specific human alcohol data exist; the reta-specificity is a community-reporting artefact.
PopulationOne human alcohol-challenge PK pilot (n=20, non-randomised); gastric-emptying/alcohol-PK physiology literature; one rat alcohol-interoception study (incl. retatrutide).
Fundingacademic / investigator-initiated (Quddos et al., alcohol-challenge pilot)
Scope limitsanimal data; human relevance uncertain; small sample (N~20)
Comparatorsnon-GLP-1 controls
GLP-1 RA classMARKETED
cns-neuropsychiatricDecrease
Low evidence
FOLK-CLAIM VERDICT ('reduces compulsive behaviours - shopping, gambling, nail-biting, skin-picking, hair-pulling'): MIXED - the food/alcohol reward base is real and...
SourceKlausen MK, Jensen ME, Moller M et al., JCI Insight, 2022Source
Full findingFOLK-CLAIM VERDICT ('reduces compulsive behaviours - shopping, gambling, nail-biting, skin-picking, hair-pulling'): MIXED - the food/alcohol reward base is real and mechanistically grounded (MATCHED to D5), but the GENERALISATION to non-consummatory compulsions (gambling, shopping, BFRBs) is UNPROVEN: zero RCTs, social-media/self-report only, plausible shared mesolimbic substrate but not demonstrated. This is belief running ahead of the evidence - and there is a counter-pole (a semaglutide-associated manic-episode case report), so the neuropsychiatric effect is not uniformly 'calming'.
PopulationExenatide AUD RCT (reward base) + a Reddit/social-media compulsive-behaviour study + a skin-picking pharmacology review (GLP-1 absent); no controlled trial in gambling/shopping/BFRBs.
Fundingacademic/independent analysis (manufacturer not study sponsor; per-study COI not individually audited)
Scope limitssmall sample (N~127)
Comparatorsplacebo
SemaglutideMARKETED
cns-neuropsychiatricNot directional
Moderate evidence
FOLK-CLAIM VERDICT ('vivid/bizarre Ozempic dreams'): UNPROVEN / essentially anecdotal. There is no randomised-trial endpoint, no sleep-architecture (PSG) study, and...
SourceWadden TA et al., JAMA Intern Med 2024 (STEP 1/2/3/5 psychiatric post hoc)Source
Full findingFOLK-CLAIM VERDICT ('vivid/bizarre Ozempic dreams'): UNPROVEN / essentially anecdotal. There is no randomised-trial endpoint, no sleep-architecture (PSG) study, and only a sub-0.1/100-patient-year uncoded spontaneous-report trickle with an explicit no-causation caveat; post-marketing spontaneous-report screening is dream-silent. More parsimonious benign explanations (REM rebound from improved sleep/OSA resolution, nocturnal hypoglycaemia in the diabetic subset, recall/attention bias and social contagion around a viral named phenomenon) are unproven but more likely than a direct dream-generating drug effect.
PopulationPooled STEP 1/2/3/5 semaglutide psychiatric-safety post-hoc analysis (Wadden 2024, the anchor source); corroborated by a post-marketing spontaneous-report screen and the manufacturer spontaneous-report rate.
Fundingindustry-Novo Nordisk
Scope limitspost-hoc (not prespecified)
Comparatorsplacebo
GLP-1 RA classMARKETED
cns-neuropsychiatricDecrease
High evidence
FOLK-CLAIM VERDICT ('smoking/nicotine cessation or reduction'): UNPROVEN-but-PROMISING (do not call it TRUE). One small positive pilot RCT (exenatide) + a consistent...
SourceYammine L, Green CE, Kosten TR et al., Nicotine Tob Res, 2021Source
Full findingFOLK-CLAIM VERDICT ('smoking/nicotine cessation or reduction'): UNPROVEN-but-PROMISING (do not call it TRUE). One small positive pilot RCT (exenatide) + a consistent semaglutide EHR signal + preclinical nicotine-reward data make it biologically coherent and encouraging, but there is no adequately powered confirmatory trial (one is ongoing). Likely a class effect but unproven as such; the EHR data are observational (no actual smoking-status data).
Fundingacademic / NIH - UTHealth Center for Clinical and Translational Sciences (NIH CTSA) + PARTNERS Research Awards; no industry sponsor (prior 'AstraZeneca' inference was WRONG)
Scope limitssmall sample (N~84)
Comparatorsplacebo; nicotine patch alone
ExenatideMARKETED
cns-neuropsychiatricMixed
Low evidence
ADDICTION (alcohol) - the counter-pole RCT: in treatment-seeking AUD patients, once-weekly exenatide added to CBT did NOT significantly reduce heavy drinking days...
SourceKlausen MK, Jensen ME, Moller M et al., JCI Insight, 2022Source
Full findingADDICTION (alcohol) - the counter-pole RCT: in treatment-seeking AUD patients, once-weekly exenatide added to CBT did NOT significantly reduce heavy drinking days (primary endpoint NEGATIVE), but attenuated fMRI alcohol-cue reactivity in the ventral striatum and lowered dopamine-transporter availability, with an exploratory benefit in the obese subgroup. The cleaner test of real-world drinking, and it failed its primary.
Fundingacademic/independent analysis (manufacturer not study sponsor; per-study COI not individually audited)
Scope limitssmall sample (N~127)
Comparatorsplacebo
ExenatideMARKETED
cns-neuropsychiatricDecrease
High evidence
ADDICTION (tobacco) - the smoking RCT: extended-release exenatide added to nicotine patch improved biochemically-confirmed smoking abstinence vs patch alone at 6 weeks...
SourceYammine L, Green CE, Kosten TR et al., Nicotine Tob Res, 2021Source
Full findingADDICTION (tobacco) - the smoking RCT: extended-release exenatide added to nicotine patch improved biochemically-confirmed smoking abstinence vs patch alone at 6 weeks and reduced post-cessation weight gain, in prediabetic/overweight smokers.
PopulationYammine pilot RCT: N=84 prediabetic/overweight smokers; double-blind; exenatide 2 mg weekly vs placebo, both with nicotine patch; 6 weeks; CO-confirmed.
Fundingacademic / NIH - UTHealth Center for Clinical and Translational Sciences (NIH CTSA) + PARTNERS Research Awards; no industry sponsor (prior 'AstraZeneca' inference was WRONG)
Scope limitssmall sample (N~84)
Comparatorsplacebo (with nicotine patch in both arms)
ExenatideMARKETED
cns-neuropsychiatricDecrease
Moderate evidence
NEURODEGENERATION (Parkinson's) - the positive phase-2: in a single-centre phase-2 trial in moderate Parkinson's, exenatide improved off-medication motor scores vs...
SourceAthauda D, Maclagan K, Skene SS et al., Lancet, 2017Source
Full findingNEURODEGENERATION (Parkinson's) - the positive phase-2: in a single-centre phase-2 trial in moderate Parkinson's, exenatide improved off-medication motor scores vs placebo at 60 weeks; the authors could not distinguish disease modification from a long-lasting symptomatic effect. The result that drove a decade of GLP-1-in-PD optimism.
Fundingacademic / philanthropic - Michael J Fox Foundation for Parkinson's Research (not AstraZeneca)
Scope limitssingle-centre/referral-enriched; small sample (N~62)
Comparatorsplacebo
ExenatideMARKETED
cns-neuropsychiatricNo change
Moderate evidence
NEURODEGENERATION (Parkinson's) - THE LOAD-BEARING NEGATIVE: the definitive multicentre phase-3 trial found that once-weekly exenatide did NOT slow Parkinson's...
SourceVijiaratnam N, Girges C, Auld G et al., Lancet, 2025Source
Full findingNEURODEGENERATION (Parkinson's) - THE LOAD-BEARING NEGATIVE: the definitive multicentre phase-3 trial found that once-weekly exenatide did NOT slow Parkinson's progression. The authors concluded there is no evidence to support exenatide as disease-modifying. This OVERTURNS the positive phase-2.
PopulationVijiaratnam/Foltynie phase-3 (NCT04232969): N=194 PD; six UK hospitals; double-blind RCT; extended-release exenatide 2 mg weekly, 96 weeks.
NEURODEGENERATION (Alzheimer's): the ELAD phase-2b trial of liraglutide in mild-moderate Alzheimer's MISSED its primary endpoint (cerebral glucose metabolism), with a...
SourceEdison P, Femminella GD, Ritchie C et al., Nat Med, 2025 (ELAD)Source
Full findingNEURODEGENERATION (Alzheimer's): the ELAD phase-2b trial of liraglutide in mild-moderate Alzheimer's MISSED its primary endpoint (cerebral glucose metabolism), with a surviving positive SECONDARY cognitive signal (ADAS-Executive) on an unadjusted p-value among several null secondaries. A missed-primary trial with a hypothesis-generating cognitive secondary, not a clean positive.
NEURODEGENERATION (Parkinson's) - the middle pole: in the LIXIPARK phase-2 trial in early Parkinson's, lixisenatide produced less motor-disability progression than...
SourceMeissner WG, Remy P, Giordana C et al., N Engl J Med, 2024 (LIXIPARK)Source
Full findingNEURODEGENERATION (Parkinson's) - the middle pole: in the LIXIPARK phase-2 trial in early Parkinson's, lixisenatide produced less motor-disability progression than placebo at 12 months (~3-point MDS-UPDRS difference), but the effect attenuated after washout and came with substantial GI side effects.
Fundingacademic / non-commercial (LIXIPARK; French PHRC government programme; disclosed)
Scope limitssmall sample (N~156)
Comparatorsplacebo
MonlunabantMARKETED
cns-neuropsychiatricMixed
Moderate evidence
CB1 CONTRAST (non-incretin landscape context): monlunabant, a CB1 inverse agonist designed to be peripherally restricted, showed dose-dependent NEUROPSYCHIATRIC...
SourceKnop FK et al., Lancet Diabetes Endocrinol, 2025 (monlunabant phase-2a, PMID 41038215); c...Source
Full findingCB1 CONTRAST (non-incretin landscape context): monlunabant, a CB1 inverse agonist designed to be peripherally restricted, showed dose-dependent NEUROPSYCHIATRIC adverse events in its phase-2a obesity trial, against the rimonabant precedent (CB1-antagonist depression/suicidality led to withdrawal). A preclinical study found monlunabant suppresses appetite via CENTRAL CB1 receptors, suggesting it is not fully peripherally confined and may carry rimonabant-like psychiatric risk. Decision-relevant CNS-safety contrast: CB1-antagonism carries a mechanism-specific neuropsychiatric liability the incretins do not appear to share.
Scope limitsconference/abstract-level; small sample (N~242)
Comparatorsplacebo
RetatrutideINVESTIGATIONAL
cns-neuropsychiatricNot directional
Moderate evidence
RETATRUTIDE GAP: retatrutide has essentially NO dedicated CNS, addiction, cognition or psychiatric data beyond routine adverse-event capture in its two phase-2 trials....
SourceJastreboff AM et al. (retatrutide phase 2), NEJM, 2023;389:514-526Source
Full findingRETATRUTIDE GAP: retatrutide has essentially NO dedicated CNS, addiction, cognition or psychiatric data beyond routine adverse-event capture in its two phase-2 trials. The absence is the finding. Any class CNS effect attributed to retatrutide is a class-transfer expectation, not a reta result - and reta's GLUCAGON arm (central glucagon/energy-balance/mood circuitry) is a CNS-relevant differentiator that is entirely unstudied, so even the class transfer may not capture it. Central glucagon circuitry intersects mood/satiety/stress pathways and could move neuropsychiatric outcomes in EITHER direction - the sign is UNKNOWN, not presumed benign. Routine adverse-event capture in the phase-2 trials is NOT a psychiatric-safety clearance for retatrutide, only an absence of dedicated data.
PopulationReta phase-2 obesity (Jastreboff, NCT04881760, N=338) + T2D (Rosenstock, NCT04867785, N=281); psychiatric AEs captured only as routine treatment-emergent events, not instrumented.
In two fully-powered phase-3 trials, oral semaglutide did NOT slow clinical progression of early Alzheimer's disease versus placebo (primary endpoint CDR-SB) - a...
SourceCummings et al., Oral semaglutide in early symptomatic Alzheimer's disease (evoke and evo...Source
Full findingIn two fully-powered phase-3 trials, oral semaglutide did NOT slow clinical progression of early Alzheimer's disease versus placebo (primary endpoint CDR-SB) - a measured null that reverses the optimistic epidemiological GLP-1/dementia-association literature.
Populationevoke and evoke+ (NCT04777396, NCT04777409): 3808 participants with early symptomatic Alzheimer's disease, oral semaglutide up to 14 mg once daily vs placebo; two multicentre randomised double-blind placebo-controlled phase-3 trials across 566 sites in 40 countries.
ADDICTION (alcohol) - the key RCT: in adults with alcohol use disorder, once-weekly low-dose semaglutide reduced alcohol consumed in a laboratory self-administration...
Full findingADDICTION (alcohol) - the key RCT: in adults with alcohol use disorder, once-weekly low-dose semaglutide reduced alcohol consumed in a laboratory self-administration task and reduced craving, drinks-per-drinking-day and heavy drinking versus placebo, but did NOT change overall drinking frequency. A within-trial smoker subsample also cut cigarettes/day.
PopulationHendershot AUD RCT (NCT05520775): N=48 non-treatment-seeking adults with AUD; phase-2 double-blind RCT, 9 weeks; semaglutide titrated to 1.0 mg vs placebo; single US centre.
Fundingacademic/independent analysis (manufacturer not study sponsor; per-study COI not individually audited)
Scope limitsanimal data; human relevance uncertain; small sample (N~48); surrogate/exploratory endpoint
Comparatorsplacebo
SemaglutideMARKETED
cns-neuropsychiatricDecrease
Low evidence
ADDICTION (alcohol) - the largest real-world signal: in an EHR cohort, semaglutide vs other anti-obesity medications was associated with substantially lower risk of...
SourceWang W, Volkow ND, Berger NA et al., Nat Commun, 2024 (Author Correction PMID 38890337)Source
Full findingADDICTION (alcohol) - the largest real-world signal: in an EHR cohort, semaglutide vs other anti-obesity medications was associated with substantially lower risk of incident and recurrent alcohol use disorder, consistent across subgroups and replicated in T2D. ASSOCIATION ONLY (LOW): confounding-by-indication is unresolved and the signal is NOT corroborated by the semaglutide AUD RCT, which was null on real-world drinking frequency.
PopulationWang/Xu EHR cohort: N=83,825 obesity (+598,803 T2D replication); 12-month retrospective cohort; semaglutide vs other anti-obesity meds.
Fundingacademic/independent analysis (manufacturer not study sponsor; per-study COI not individually audited)
Scope limitspossible confounding by indication
Comparatorsother anti-obesity medications
SemaglutideMARKETED
cns-neuropsychiatricDecrease
Low evidence
ADDICTION (tobacco): in a target-trial-emulation EHR study of T2D + tobacco use disorder, semaglutide was associated with lower TUD-related healthcare measures versus...
SourceWang W, Volkow ND, Berger NA et al., Ann Intern Med, 2024Source
Full findingADDICTION (tobacco): in a target-trial-emulation EHR study of T2D + tobacco use disorder, semaglutide was associated with lower TUD-related healthcare measures versus seven other antidiabetes classes, including other GLP-1RAs.
PopulationWang/Xu target-trial emulation (US EHR): N=222,942 new antidiabetes users incl. 5,967 semaglutide with T2D+TUD; 12 months; vs 7 comparator classes.
Fundingacademic/independent analysis (manufacturer not study sponsor; per-study COI not individually audited)
NEURODEGENERATION (Alzheimer's) - the load-bearing NEGATIVE readout: the EVOKE and EVOKE+ phase-3 trials of oral semaglutide in early Alzheimer's did NOT slow disease...
SourceNovo Nordisk topline announcement, 24 Nov 2025 (EVOKE/EVOKE+); CTAD Dec 2025; peer-review...Source
Full findingNEURODEGENERATION (Alzheimer's) - the load-bearing NEGATIVE readout: the EVOKE and EVOKE+ phase-3 trials of oral semaglutide in early Alzheimer's did NOT slow disease progression (CDR-SB primary not met) despite company-reported improvement in AD biomarkers - a biomarker/clinical dissociation. The largest GLP-1-in-AD result to date.
PopulationEVOKE (NCT04777396) + EVOKE+ (NCT04777409): amyloid-positive early AD (MCI / mild dementia), age 55-85; phase-3 double-blind placebo-controlled; ~2 years.
Fundingindustry - Novo Nordisk (trial sponsor; inferred from registration trial)
Scope limitsidentifier not fully verified; no outcome data yet (ongoing)
The Echo Compendium