GLP-1 And Addiction: Mechanism, Missteps, And What Comes Next

By Jason Kirby, chief medical officer, Recovery Centers of America

A growing body of evidence suggests glucagon-like peptide-1 receptor (GLP-1R) agonists may influence reward pathways and craving, raising important questions for addiction research and drug development. GLP-1R agonists were originally engineered to regulate incretin signaling, improve glycemic control, and slow gastric emptying.¹ They were not designed to influence drug-seeking behavior — yet preclinical models and early human observations suggest effects on reward processing and craving. A rapidly accumulating body of preclinical data and early clinical observations suggests that GLP-1R signaling extends well beyond its metabolic origins, with emerging relevance in the neurobiology of addiction.²

For researchers and developers working in CNS drug discovery, this signal creates an obligation: interrogate the mechanism rigorously, hold clinical translation to a high evidentiary standard, and ask what this actually tells us about reward neurobiology and where it could take the pipeline.

Receptor biology is where the interrogation begins.

An Unexpected Entry Point Into The Brain’s Reward Circuitry

In CNS drug development, mechanism matters — but location determines impact.

GLP-1R expression was initially characterized in the hypothalamus, brainstem, and pancreas. That is the expected architecture for a hormone coordinating energy intake and insulin secretion. The finding that reframes its relevance to addiction is that GLP-1R is also expressed in the ventral tegmental area (VTA), nucleus accumbens (NAc), and prefrontal cortex (PFC). These are not peripheral metabolic nodes; they are the core structures of the mesocorticolimbic dopamine system,³ the circuitry that addiction medicine has spent decades trying to modulate pharmacologically.

Translational confidence was initially limited by species-specific differences in GLP-1R expression patterns. A more recent postmortem human brain mapping study,⁴ using RT-PCR and immunohistochemistry across 30 samples, confirmed GLP-1R presence in mesocorticolimbic regions, including the frontal cortex, PFC, hippocampus, and thalamus. Frontal cortical expression appears proportionally more abundant in humans than in rodent models. That difference has direct implications for how we interpret behavioral data generated in standard preclinical systems.

If the human GLP-1R expression profile differs from rodent models in both magnitude and regional distribution, standard preclinical endpoints cannot be assumed to predict clinical outcomes. That scrutiny is exactly where early-stage development teams should invest analytical effort.

Distribution alone is not enough, though. The next question is how GLP-1R activation alters signaling within those circuits. Early-stage programs should prioritize human tissue validation early, rather than relying on rodent expression patterns as a proxy for target engagement.

Mechanism Of Action: A More Targeted Signal

Receptor location explains possibility. Mechanism determines relevance.

GLP-1R belongs to the class B G protein-coupled receptor (GPCR) family.⁵ It signals primarily through Gαs to elevate cAMP and activate PKA, while also engaging β-arrestin-mediated MAPK cascades. The balance between these arms (what the field calls biased agonism) varies meaningfully across compounds. It has direct implications for CNS drug development.

Within mesocorticolimbic circuitry, GLP-1R activation does not function as a blunt dopamine antagonist. The preclinical evidence points to something more selective: context-dependent recalibration of phasic dopamine release rather than tonic suppression of the dopaminergic system.⁶ In alcohol models, GLP-1R agonist administration attenuates ethanol-induced dopamine overflow in the NAc without disrupting baseline dopaminergic tone.⁷ In psychostimulant models, the primary mechanism involves modulation of dopamine transporter function rather than global receptor blockade.⁸ In both cases, the behavioral result is reduced reward salience without the emotional blunting that has historically plagued dopamine-targeting pharmacotherapies.

At the synaptic level, central administration of exendin-4 increases AMPA/kainate receptor-mediated excitatory postsynaptic currents in VTA dopamine neurons and reduces paired pulse ratios.⁹ This indicates enhanced presynaptic glutamate release. In the NAc core, GLP-1R activation suppresses reward-driven consumption through a glutamatergic pathway reversed by AMPA/kainate antagonism. The implication is clear: anti-reward effects are mediated through excitatory circuit modulation, not direct dopaminergic inhibition.

The gut–brain axis adds a third mechanistic thread. GLP-1-producing neurons in the nucleus tractus solitarius (NTS) project directly to the VTA and NAc,¹⁰ establishing a physiological route through which peripheral metabolic signals impact central reward processing. Vagotomy abolishes the effects of peripherally administered GLP-1R agonists on reward behavior. That single finding has direct patient selection implications. Diminished vagal tone is common in populations with chronic alcohol or opioid use disorder, and it may substantially limit peripheral compound efficacy in those patients.

Development programs should treat metabolic state and vagal integrity as testable variables, not background noise, and incorporate them into early pharmacodynamic modeling. These mechanisms are now being directly tested in preclinical models.

Preclinical Evidence: Consistent Signal With Caveats

The preclinical signal is consistent, which makes careful interrogation more important. Across multiple substance classes and rodent models, GLP-1R agonists reduce self-administration; conditioned place preference; and cue-induced reinstatement for alcohol, nicotine, cocaine, and opioids.¹¹ Behavioral suppression is reproduced with direct central infusion into the VTA and NAc, establishing neuroanatomical specificity. This is a more coherent evidence base than addiction neuroscience typically generates at this stage of target development.

However, three caveats need to be considered:

1. Species expression problem. Quantitative and regional differences in human versus rodent GLP-1R distribution mean that circuit-level effects observed in mice may not map cleanly onto human neurobiology, even if the underlying pharmacology is conserved.
2. Metabolic confound. Most preclinical studies use animals with experimentally controlled metabolic states that do not reflect the heterogeneity of real addiction populations. Human data suggest GLP-1R agonist effects on alcohol consumption may be substantially larger in individuals with comorbid obesity.¹² One randomized controlled trial found increased alcohol intake in lean participants following exenatide treatment.¹³ If the anti-reward mechanism depends on metabolic pathways only engaged under conditions of metabolic dysregulation, the patient selection logic for any addiction indication changes fundamentally.
3. Model validity. Rodent self-administration paradigms capture important features of compulsive drug use. They do not model the psychological, social, and environmental complexity of human substance use disorders, but early clinical observations suggest that similar patterns may be emerging in humans.

Preclinical programs should explicitly model metabolic variability and avoid overreliance on single-species behavioral endpoints when advancing compounds into early clinical testing.

Best Practices For Early-Stage Development

Several principles emerge from the current evidence base that should guide early-stage development programs:

• Prioritize human-relevant validation early. Species-specific differences in GLP-1R distribution necessitate early incorporation of human tissue data to confirm target engagement beyond rodent proxies.
• Stratify by metabolic phenotype. Evidence suggests treatment effects may differ significantly between metabolically healthy and dysregulated populations, making patient selection a primary design variable rather than a secondary consideration.
• Design trials to prospectively test subgroup effects. Post-hoc findings in obesity-linked cohorts should be treated as hypotheses and validated through prespecified stratification.
• Align endpoints with clinical relevance. Laboratory self-administration and craving measures should be complemented with endpoints that reflect real-world behavior change and recovery outcomes.
• Integrate physiological moderators into pharmacodynamic models. Variables such as vagal tone, heart rate variability, and metabolic state should be incorporated into early modeling rather than treated as noise.

Common Pitfalls To Avoid

Despite a consistent preclinical signal, several recurring missteps could compromise translation:

• Overreliance on rodent behavioral models. These models capture reinforcement but fail to reflect the psychosocial complexity of human substance use disorders.
• Assuming translatability of receptor distribution. Differences in regional expression between species may distort expectations of clinical efficacy.
• Overinterpreting subgroup analyses. Positive signals in exploratory analyses, particularly those tied to metabolic status, require prospective validation before informing development strategy.
• Treating metabolic effects as independent from reward modulation. The interaction between metabolic physiology and reward circuitry appears central to mechanism and should not be compartmentalized.
• Advancing compounds without clear CNS pharmacodynamic understanding. Variability in blood–brain barrier penetration and signaling bias can meaningfully alter both efficacy and safety profiles.

Early Clinical Observations

The clinical data is limited, easily overinterpreted, and largely derived from populations enrolled for metabolic rather than addiction endpoints. That context is essential.

For alcohol use disorder, a double-blind placebo-controlled trial with exenatide found no significant reduction in heavy drinking days in the overall cohort.¹³ That null primary endpoint is frequently underemphasized in secondary discussions of the data. A positive signal emerged only in an exploratory subgroup analysis restricted to participants with obesity, making it hypothesis generating rather than confirmatory. A more recent Phase 2 trial¹⁴ with low-dose semaglutide showed reductions in laboratory-based alcohol self-administration and craving scores, with some effect sizes exceeding Cohen’s D of 0.80. While encouraging, laboratory self-administration is only a proximal endpoint. It does not translate directly to a real-world recovery measure.

For opioid use disorder, preliminary data from a small randomized trial reported a 40% reduction in craving with liraglutide following presentation at a 2023 scientific conference.¹⁵ The result has not yet appeared in peer-reviewed publication. The sample is small, the duration is short, and the findings remain unpublished. Replication in a powered addiction-primary trial is the necessary next step.

For stimulant use disorder, a single low dose of exenatide failed to reduce craving or euphoria in a cocaine use disorder cohort, despite producing measurable hormonal changes.¹⁶ Whether the dose was insufficient, the duration too short, or metabolic phenotype stratification essential remains unclear. Ongoing trials are exploring repeated dosing strategies. That is the right scientific response, but it also illustrates how much foundational dose finding and pharmacodynamic characterization lies ahead.

While those early signals are intriguing, they also clarify how many development questions are unresolved. Early clinical programs should be designed to replicate subgroup signals prospectively, rather than relying on post-hoc analyses to guide development decisions.

Key Development Challenges

The key issue is not whether the signal exists but whether it can be developed responsibly. Several specific gaps need to be addressed before GLP-1R agonism can be reliably advanced toward late-phase addiction development:

1. CNS penetrance and neuropsychiatric risk. GLP-1R agonist compounds vary substantially in blood–brain barrier penetrance. Greater CNS penetrance produces stronger central pharmacodynamic effects but also higher rates of neuropsychiatric adverse events, including worsening depression, anxiety, apathy, and rare suicidality signals. In addiction populations, where psychiatric comorbidity rates are high,¹⁷ this is not a peripheral safety concern. It requires prospective monitoring in every early-phase trial and should inform compound selection upstream.
2. Biased agonism. The Gαs-versus-β-arrestin balance differs across compounds and influences receptor desensitization kinetics and downstream transcriptional programs. The β-arrestin/MAPK arm regulates gene expression and long-term cellular plasticity, both potentially relevant to the neuroadaptations underlying cravings and relapse. Whether Gαs-biased agonism is advantageous for addiction circuit applications is not currently known, and it should not be assumed.
3. Metabolic state as a moderator. Vagal tone, baroreflex sensitivity, and heart rate variability should be integrated into early-phase trial designs as stratification variables and pharmacodynamic endpoints. The evidence that metabolic phenotype moderates treatment response is a proposed mechanism. It demands prospective testing.
4. Endpoint heterogeneity. Laboratory self-administration tasks, craving scales, and drinking diary data measure different constructs. They respond differently to pharmacological intervention. Validated addiction-specific endpoints that capture clinically meaningful outcomes are a prerequisite for credible late-phase development.

Taken together, these challenges point toward a more selective and mechanistically disciplined development strategy. These variables should be built into trial design up front, not addressed retrospectively once variability appears in the data.

What A Rational Development Strategy Looks Like

A rational development strategy starts with constraint, not expansion. Initial addiction studies should probably enrich for individuals with comorbid metabolic dysregulation because target engagement is most likely robust there and pharmacodynamic hypotheses can be tested cleanly. Understanding who responds and why is more valuable at this stage than generating broad efficacy data in heterogeneous cohorts.

A CNS-selective GLP-1R agonist with a tailored signaling bias profile optimized for mesocorticolimbic modulation rather than metabolic endpoints is a more scientifically principled development candidate than repurposing existing metabolic agents. Combination strategies that enhance natural GLP-1 signaling through gut-targeted mechanisms, such as DPP-4 inhibition, L-cell stimulation, or transcutaneous vagal neuromodulation, could preserve central efficacy while reducing CNS adverse event burden. A more favorable therapeutic window may be achievable.

The most important possibility here extends beyond repurposing. GLP-1R signaling has revealed a previously underappreciated interface between metabolic physiology and reward neuroscience. That interface could generate novel mechanistic insights and purpose-built therapeutic candidates. Getting there requires the same rigor and mechanistic discipline that any credible CNS program demands.

Implications For Treatment Models And Clinical Integration

If validated, GLP-1R agonism may require a more integrated treatment framework than traditional addiction pharmacotherapies. The apparent interaction between metabolic state and reward modulation suggests that treatment models may need to incorporate metabolic assessment as part of addiction care planning.

For providers, this could mean identifying subpopulations — such as individuals with comorbid obesity or metabolic dysregulation — where therapeutic response is more likely. For developers, it reinforces the need to design therapies that account for both central and peripheral physiology.

Rather than functioning as a stand-alone intervention, GLP-1-based approaches may ultimately integrate into broader multimodal treatment strategies that address both neurobiological and systemic drivers of substance use.

The signal justifies further investment. It does not justify shortcuts. Early programs should prioritize signal clarity over speed, even if that delays broader indication expansion.

References

1. Science Direct – Glucagon-Like Peptide-1 Receptor Agonists in Chronic Kidney Disease: Mechanisms and Clinical Perspectives
2. National Library of Medicine – Mechanisms of GLP-1 in Modulating Craving and Addiction: Neurobiological and Translational Insights
3. National Library of Medicine – Anatomy and Function of Ventral Tegmental Area Glutamate Neurons
4. Medical Sciences – Mechanisms of GLP-1 in Modulating Craving and Addiction: Neurobiological and Translational Insights
5. PubMed Central – Glucagon-Like Peptide-1 and Its Class B G Protein-Coupled Receptors: A Long March to Therapeutic Successes
6. PubMed Central – Central GLP-1 receptor activation modulates cocaine-evoked phasic dopamine signaling in the nucleus accumbens core
7. National Library of Medicine – Effects of GLP-1 Receptor Agonists in Alcohol Use Disorder
8. Frontiers – The potential role of GLP-1 receptor agonists in substance use disorders – a systematic review
9. PubMed Central – The food intake-suppressive effects of glucagon-like peptide-1 receptor signaling in the ventral tegmental area are mediated by AMPA/kainate receptors
10. PubMed Central – GLP-1 Neurons in the Nucleus of the Solitary Tract Project Directly to the Ventral Tegmental Area and Nucleus Accumbens to Control for Food Intake
11. Science Direct – IUPHAR review – Glucagon-like peptide-1 (GLP-1) and substance use disorders: An emerging pharmacotherapeutic target
12. Science Direct – Effects of glucagon-like peptide-1 receptor agonists on alcohol consumption: a systematic review and meta-analysis
13. National Library of Medicine – Exenatide once weekly for alcohol use disorder investigated in a randomized, placebo-controlled clinical trial
14. JAMA Psychiatry – Once-Weekly Semaglutide in Adults With Alcohol Use Disorder
15. STAT – Opioid cravings were reduced by anti-obesity drug in small study
16. National Library of Medicine – Testing the effects of the GLP-1 receptor agonist exenatide on cocaine self-administration and subjective responses in humans with cocaine use disorder
17. National Library of Medicine – Common Comorbidities with Substance Use Disorders Research Report

About The Author

Jason Kirby, DO, MBA, DFASAM, is chief medical officer at Recovery Centers of America where he leads enterprisewide clinical strategy, quality improvement, and outcomes-driven care across residential, outpatient, and virtual settings. He is dual board-certified in addiction medicine and family medicine, with more than a decade of clinical and executive experience in addiction treatment and behavioral health.

Kirby is a Distinguished Fellow of The American Society of Addiction Medicine and serves as treasurer on their National Executive Board. He is a 2024 – 2025 Health Policy Fellow with the American Association of Osteopathic Medicine, with a focus on healthcare policy and regulatory strategy in addiction treatment. He earned his DO from West Virginia School of Osteopathic Medicine and his MBA from Point Park University.