A systematic review on the role of glucagon-like peptide-1 receptor agonists on alcohol-related behaviors: potential therapeutic strategy for alcohol use disorder

This systematic review was conducted in adherence to the 2020 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (Page et al., 2021). Relevant literature was extracted from the following databases: Web of Science, OVID (MedLINE, Embase, AMED, PsychInfo, JBI EBP), and PubMed were used to systematically search for relevant articles from database inception to October 27th, 2024. The search queries used for this review is of the following: (‘GLP-1’ OR ‘Glucagon-Like Peptide-1’ OR ‘Glucagon-Like Peptide 1’ OR ‘GLP-1 Agonist’ OR ‘Glucagon-Like Peptide-1 Agonist’ OR ‘Glucagon-Like Peptide 1 Agonist’ OR ‘Semaglutide’ OR ‘Dulaglutide’ OR ‘Trulicity’ OR ‘Exenatide’ OR ‘Liraglutide’ OR ‘Lixisenatide’ OR ‘Tirzepatide’) AND (‘Alcohol’ OR ‘Alcohol use disorder’ OR ‘AUD’ OR ‘Alcoholism’ OR ‘Ethanol Administration’ OR ‘Ethanol’).

Articles obtained from databases were screened through Covidence, wherein duplicated articles were automatically excluded (Systematic review management, 2023). Two independent reviewers (H.A. and Y.J.Z.) reviewed the titles and abstracts according to the inclusion and exclusion guidelines (Table 1). Furthermore, relevant English-language primary and secondary sources were then examined through full-text screening and included for extraction if they reported on GLP-1RA-associated changes to alcohol or ethanol consumption. Discrepancies were resolved following thorough discussion between reviewers.

Literature applicable to alcohol-associated behaviours and GLP-1RA administration were obtained and further organised following the piloted data extraction guidelines (Table 2). Extracted data was established a priori, involving (1) author(s), (2) study design, (3) Sample Size, (4) measurement tools, and (5) outcome of interest(s) for preclinical studies (Table 2). An additional category, diagnosis, was included for clinical studies (Table 3). Two reviewers (C.S. and Y.J.Z.) extracted data from relevant literature, followed up with thorough discussion to resolve potential conflicts. Alcohol consumption and outcome of interest(s) involved GLP-1 or GLP-1RA administration in the context of alcohol-related behaviours in preclinical and clinical literature.

Quality assessment of pre-clinical studies was conducted using SYRCLE’s risk of bias analysis tool for animal studies, whereas quality assessment of clinical studies was conducted using Quality Assessment of Controlled Intervention Studies of the National Institute of Health (Hooijmans et al., 2014; Ma et al., 2020; National Institute of Health, 2013). Similarly, quality assessment of cohort studies was conducted using the Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies (Ma et al., 2020; National Institute of Health, 2013). Independent reviewers (H.A., C.S., and Y.J.Z.) examined literature risk of bias and resolved all conflicts through further discussions. These tools were chosen for their established validity and reliability in assessing risk of bias in animal, clinical, and observational studies. The inclusion and exclusion criteria, comprehensive overview of data, and methodological quality are organised as supplementary materials (Table S1↗, S2↗).

A systematic search of relevant literature yielded 1,309 studies. Among these, 31 duplicates were identified manually, and 590 were identified by Covidence (Fig. 1). 688 studies were screened based on their titles and abstracts according to the inclusion and exclusion criteria (Table 1). Following this, 28 full-text studies were assessed, with 21 studies deemed relevant and included for data extraction (Table 2). 6 studies were excluded, consisting of studies with incorrect study designs (n = 4), without available full text (n = 1), and with wrong outcomes (n = 1). Ultimately, 21 studies meeting the eligibility criteria were included in this systematic review (Table 2).

All controlled intervention studies were described as double-blinded, randomised trials, and had sufficient sample size to detect a clinically significant effect. Included observational cohort studies had sufficient sample size to detect a clinically significant effect, whilst employing consistent exposure measures across all study participants. Notwithstanding, Quddos et al. (2023) did not clearly specify the study population, and did not employ random assessment (Table S2↗). As such, we only examined the objective results investigating the association between GLP-1 RA administration and alcohol consumption behaviour for this study.

Most animal studies employed similar groups at baseline and/or adjusted for confounders and randomly housed animals during the experiment. Common limitations included the lack of blinding and random outcome assessment. Most studies did not exhibit a high degree of bias. Notwithstanding, studies rated ‘fair’ generally exhibited higher levels of categories marked as ‘Not Reported’ or ‘Not Applicable’, further omitting information regarding allocation concealment and insufficiently addressing incomplete outcome data (Table S1↗). Albeit the presence of limitations, the qualities of results were not affected.

Cumulatively, we identified 21 studies investigating the role of GLP-1RA on alcohol consumption or administration in preclinical (n = 19) and clinical (n = 2) studies. The GLP-1RAs identified and assessed in this review are exenatide, semaglutide, and liraglutide.

To understand the influence of GLP-1RAs (i.e., exenatide, semaglutide, and liraglutide) on alcohol consumption, 19 animal studies, consisting of rodents and monkey models, were identified. These studies used intracerebral injections to target varying brain regions, as well as peripheral infusion, to assess alcohol-associated behaviours.

We identified 13 studies evaluating the impact of exenatide on alcohol consumption, demonstrating that exendin-4 (Ex-4) administration attenuates alcohol-related behaviours across various brain regions (Colvin et al., 2020; Allingbjerg et al., 2023; Díaz-Megido & Thomsen, 2023; Egecioglu et al., 2013; Dixon et al., 2020; Shirazi et al. 2013; Sørensen et al., 2016; Suchankova et al., 2015; Thomsen et al., 2017; Thomsen et al., 2018; Vallof et al., 2019 a; Vallof et al., 2019 b).

Allingbjerg et al. (2023) administered Ex-4 at varying doses (i.e., vehicle, 3.2, 10, 32 ng/hemisphere) in the NAc, ventral hippocampus (HPC), and lateral septum (LS), resulting in attenuated alcohol self-administration in C57BL/6J mice with a substantial effect size. Specifically, when micro-infused into the ventral HPC, Ex-4 significantly decreased self-administration compared to the vehicle group (p < .0001). Similar reductions were observed following administration to the NAc, F (3, 21) = 4.46, p = 0.03, and lateral septum (LS), F (3, 21) = 8.17, p = 0.001. Post hoc analyses further supported that doses of 0.01 and 0.03 ng/hemisphere in both the NAc and LS significantly reduced alcohol self-administration compared to vehicle (NAc: p = 0.047 and p = 0.049; LS: p = 0.03 and p = 0.006). A higher dose of Ex-4 (4.8 μg/kg) significantly reduced alcohol-induced locomotor stimulation and attenuated alcohol-induced dopamine release in the NAc (P < 0.01 to P < 0.001), highlighting its impact on both behavioural and neurochemical responses to alcohol (Egecioglu et al., 2013).

Additional findings by Colvin et al. (2020), support that Ex-4 administration into the ventral tegmental area (VTA) significantly reduced alcohol intake at 0.05 µg and 0.5 µg (F (3,21) = 50.7, p < 0.0001), aligning with reductions observed in the NAc core and shell (NAcC: F (3,21) = 37.7, p < 0.001; NAcS: F (3,21) = 34.6, p < 0.001). In the Dorsomedial Hippocampus (DMHipp), only the highest dose of 0.5 µg produced a significant reduction in alcohol consumption F (3,21) = 22.1, p < 0.01), indicating a dose-dependent effect specific to this region.

To further support these findings, Dixon et al. (2020) demonstrated that intra-VTA injections of Ex-4 (i.e., vehicle, 0.01 μg, 0.05 μg) yielded reduction in alcohol-self administration, which was most prominent in high-alcohol Long-Evans rats. It appears that Ex-4 selectively reduced alcohol-seeking behaviour in rats with a demonstrated specificity toward high-alcohol-drinking (HAD) phenotypes. In Experiment 1, Ex-4 treatment significantly attenuated non-reinforced nose pokes (F (2,42) = 4.206, p < 0.05), alcohol infusions (F (2,42) = 4.424, p < 0.05), and total alcohol consumption (F (2,40) = 4.181, p < 0.05), indicating that active nose pokes were directly associated with alcohol consumption rather than non-specific increases in activity. However, in Experiment 2, Ex-4 had no impact on reacquisition behaviours, as evidenced by unchanged alcohol deliveries (t (21) = 0.773, p > 0.05), and in Experiment 3, Ex-4 administration in the VTA showed no effect on motivation to seek alcohol (t (21) = 0.666, p > 0.05). This lack of effect on reacquisition and motivation suggests that Ex-4’s influence is not on general alcohol-seeking motivation but rather specific to conditions of established high alcohol intake.

Likewise, in high alcohol-consuming rats, 1.2 μg/kg Ex-4 effectively reduced alcohol intake across multiple time points ([1 hour: F (2,19) = 18.69, P < 0.001; 4 hours: F (2,19) = 12.95, P < 0.001; 24 hours: F (2,19) = 8.236, P < 0.01]), whereas a lower 0.03 μg/kg dose had no significant effect (Egecioglu et al., 2013). Additionally, in an operant self-administration test, 1.2 μg/kg Ex-4 significantly reduced active lever presses (P < 0.01) in rats with prolonged alcohol exposure, indicating its potential to suppress established alcohol-seeking behaviours.

In conditioned place preference (CPP) experiments, a 2.4 μg/kg dose of Ex-4 significantly reduced alcohol-induced CPP (P < 0.01) without affecting CPP in vehicle-conditioned mice (P > 0.05), demonstrating specificity for alcohol-related cues. When administered throughout conditioning, Ex-4 inhibited alcohol’s effect on CPP (P < 0.05), and co-treatment in non-alcohol-conditioned mice did not induce CPP (P > 0.05). An increased dose (i.e., 3.2 μg/kg Ex-4) further reduced alcohol consumption relative to baseline (P < 0.01) (Sørensen et al., 2016). Furthermore, mice response to alcohol reinforcement was reduced to 16% of the baseline, demonstrating reduction in alcohol-seeking behaviours.

Further evidence in accordance with the aforementioned findings are provided by Vallof et al. (2019 a), wherein researchers infused Ex-4 into the NAcS in low and high alcohol consuming mice. Ex-4 infusion into the NAcS (0.05 μg per hemisphere) resulted in an attenuation of alcohol reward memory under CPP testing (t (13) = 2.15, P = 0.0257)). However, there were no distinctions in CPP responses between control and Ex-4 mice (t (11) = 0.47, P = 0.3224)). Additionally, low alcohol consuming rats showed no significant alterations (t (5) = 0.76, P = 0.4836)) following 12-weeks of baseline alcohol consumption in Ex-4 and vehicle rats (Vallof et al., 2019 a). Locomotor response to alcohol was further inhibited by nucleus of the solitary tract (NTS)-Ex-4 administration at 0.05 μg per hemisphere at one, two, and 24 hours (P < 0.001) (Vallof et al., 2019 b). Alcohol response was significantly decreased in Ex-4 treated mice (P < 0.05, F (3,49) = 2.85, P = 0.0469). Consistent with existing literature, Ex-4 treated mice showed a declined presence for the alcohol-associated compartment compared to controls in CPP testing (P < 0.05) (Vallof et al., 2019 b).

Alongside Ex-4’s disparate impact in varying drinking phenotypes (i.e., high versus low consumption), studies further characterised potential sex differences and alcohol-related behaviours following Ex-4 administration (Díaz-Megido & Thomsen, 2023). For cue-induced reinstatement, systemic Ex-4 treatment (saline, 1.8, and 3.2 μg/kg) had no impact on reinstatement responses in female mice, who showed condition-dependent nose poking [F (2,54) = 64.6, p < 0.0001] with a lack of treatment effect (p > 0.4) (Díaz-Megido & Thomsen, 2023). Post hoc analysis confirmed that reinstatement responses exceeded extinction, and inactive responses decreased from baseline to extinction (both p < 0.0001), with no significant differences between reinstatement and extinction (p > 0.8). In male mice, however, Ex-4 significantly reduced reinstatement behaviour, showing a dose-dependent response ([F (2,16) = 4.57, p = 0.03]).

Analysis further revealed a dose effect ([F (2,43) = 5.20, p = 0.01]) and a sex-by-dose interaction ([F (2,43) = 4.87, p = 0.01]), driven by male responses. Inactive nose-poke responses were unaffected by sex or dose (p > 0.4). Additionally, Ex-4 significantly reduced oral alcohol self-administration in males ([F (2,18) = 5.80, p = 0.01]), with both doses decreasing intake relative to saline group (p = 0.01, p = 0.006). Notwithstanding, female mice had no significant dose effect (p = 0.25). Additional analysis for the first- and second-hour following treatment indicated Ex-4 reduced male self-administration in the first hour ([F (2,18) = 8.13, p = 0.003), with both doses yielding lower alcohol-related behaviours than saline (i.e., p = 0.02, p = 0.0009). Female mice displayed a non-significant trend (p = 0.08), with no significant effects observed in the second hour (p = 0.26 − 0.70) (Díaz-Megido & Thomsen, 2023).

Correspondingly, Thomsen et al. (2017) identified a significant increase in alcohol consumption and preference compared to baseline in saline (p ≤ 0.01, days 11–18 vs. baseline) but not Ex-4 C57BL/6NTac mice groups, post-alcohol deprivation. Ex-4 significantly reduced alcohol intake and preference compared to saline (p < 0.05, days 11–18). In the post-deprivation period, Ex-4 significantly reduced the number of alcohol drinking bouts ([F (1,13) = 11.6, p < 0.01]) and delayed the first alcohol drink compared to the saline group (χ² = 5.13, p < 0.05). Only one of nine Ex-4 mice chose alcohol first, versus five of six in the saline group. With Ex-4 discontinuation, both measures normalised to saline levels within a day, with no group differences during washout (days 19 to 25). Thomsen et al. (2018) followed up their research in monkeys, wherein alcohol consumption was significantly attenuated following exenatide treatment in week one of alcohol reintroduction (F (1,21) = 6.80, p = 0.02), but this effect was not preserved in week two.

With peripheral injection of Ex-4, there was also a significant reduction in alcohol consumption, without alterations in general fluid intake (Shirazi et al., 2013). Similar effects were observed in a modified version of Ex-4 (i.e., AC3174), wherein AC3173 doses significantly reduced alcohol consumption relative to vehicle EtOH mice (Suchankova et al., 2015). Prior to treatment, EtOH (alcohol-dependent) mice consumed significantly more alcohol than control mice (F (1,67) = 13.61, P < 0.0001). Following, AC3174 administration reduced alcohol drinking, but did not significantly affect alcohol consumption in non-alcohol-dependent control mice.

Literature further demonstrated that selective administration in brain regions is crucial for Ex-4 effectiveness, as certain brain regions may not be an optimal targeting site. Notably, Ex-4 did not yield significant effects when injected in the caudate putamen (CPu), basolateral amygdala (BLA), arcuate nucleus (ArcN), paraventricular nucleus (PVN), anterior VTA (aVTA), posterior VTA (pVTA), and lateral dorsal tegmental nucleus (LDTg) (Colvin et al., 2020; Allingbjerg et al., 2023; Vallof et al., 2019 a). Ex-4 administration into the CPu, representing dorsal striatal targeting, did not produce significant alterations in alcohol self-administration behaviour, likely reflecting the lower density or absence of GLP-1Rs in this region [F (3, 18) = 0.89, p = .44] (Allingbjerg et al., 2023). Additionally, Ex-4 administration did not significantly influence alcohol intake in the BLA, ArcN, or PVN (p > 0.05) (Colvin et al., 2020). Research also identified lack of significant effect on alcohol-related behaviours in the aVTA, pVTA, and LDTg, wherein Ex-4 infusion (0.0025 μg per hemisphere) had limited effects on CPP response, alcohol consumption, and alcohol preference (Vallof et al., 2019 a). In the aVTA, Ex-4 did not alter alcohol reward memory in conditioned place preference (CPP) compared to vehicle (t(13) = 0.85, P = 0.4111) and had no effect on alcohol intake (t(7) = 0.36, P = 0.7324) or preference (t(7) = 0.13, P = 0.9023) in high alcohol-consuming rats. Similarly, in the pVTA, Ex-4 did not affect CPP (t (12) = 0.61, P = 0.5503), alcohol intake (t (7) = 0.10, P = 0.9219), or preference (t (7) = 0.05, P = 0.9644) in high alcohol-consuming rats. In the LDTg, Ex-4 infusion had no effect on CPP in either low or high alcohol-consuming rats (t (44) = 0.54, P = 0.5914), with no differences in CPP between vehicle and Ex-4 groups in vehicle-conditioned rats (t (13) = 0.81, P = 0.4329). In low alcohol-consuming rats, Ex-4 had no effect on baseline alcohol intake (t (13) = 0.53, P = 0.5270) or preference (t (13) = 0.24, P = 0.8128). However, in high alcohol-consuming rats, Ex-4 significantly reduced alcohol intake (t (11) = 3.26, P = 0.0076) without affecting alcohol preference (t (11) = 1.42, P = 0.1847). These findings suggest Ex-4’s selective efficacy in reducing intake specifically in high-consuming phenotypes, with minimal effects on alcohol reward memory or preference across regions (Vallof et al., 2019 a).

We identified 4 studies that reported on the effects of semaglutide on alcohol consumption in preclinical studies (Aranas et al., 2023; Marty et al., 2020; Chuong et al., 2023; Fink-Jensen et al., 2024). Aranas et al. (2023) reported that acute and repeated semaglutide treatment significantly decreased alcohol intake in both male and female rats, with more pronounced reductions in males, suggesting a potential link between alcohol intake and body weight (P < 0.0001). Higher reductions in alcohol consumption were observed with higher dosages, indicating a dose-dependent association. Furthermore, semaglutide was also found to prevent post-withdrawal alcohol consumption (t (14) = 2.67, P = 0.0187, paired t-test, n = 15) and relapse drinking, with significant reductions reported at the 24th and 48th hour. Similarly, Chuong et al. (2023) observed that semaglutide significantly reduced preference for both sweetened and unsweetened alcohol (P < 0.0001) across all doses, but significant reductions were reported for doses 0.003mg/kg, 0.01mg/kg, 0.03mg/kg, and 0.1mg/kg. While female rodents were found to consume more alcohol, no significant dose-sex interaction was reported. Despite the general efficacy of semaglutide, Fink-Jensen et al. (2024) found that while alcohol consumption was significantly less in semaglutide-induced mice than vehicle controls (P = 0.0055), reductions only lasted through the second week (week 1: P = 0.0428, week 2: P = 0.0022, week 3: P = 0.0632). This finding suggests that the drug’s effect may be short-lived. Its transient effects were also supported by Marty et al. (2020)’s findings, as researchers found that while semaglutide did reduce ethanol consumption and preference, the effects did not last more than two days post-drug administration. Furthermore, semaglutide demonstrated the ability to selectively reduce alcohol consumption without affecting other fluid intake, and these reductions were unaffected by the administration of Ex9-39, a GLP-1R antagonist, suggesting that its effects are independent.

With respect to liraglutide treatment, four studies exploring its effect on alcohol consumption were identified (Thomsen et al., 2018; Marty et al., 2020; Liu et al., 2024; Vallof et al., 2015). These studies explore the effects liraglutide has on consumption and preference, and the complexities of its underlying mechanisms. Findings by Liu et al. (2024) reported that reductions in alcohol consumption and preference were observed in alcohol treated mice treated with liraglutide for 6 weeks in comparison to the control group. Notably, these mice continued to exhibit these effects (relative to the control group) even after a two-week alcohol withdrawal period. Interestingly, total fluid intake was unaffected in both groups, suggesting that the reduction in alcohol preference was due to a selective effect on alcohol rather than general changes in fluid intake.

Correspondingly, Marty et al. (2020)’s findings demonstrated that while liraglutide did reduce ethanol intake and preference (liraglutide: treatment × time-point interaction), the effects were transient and did not last for more than 2 days post-injection. The researchers proposed that, forasmuch as liraglutide may have increased water intake, it accounts for the reduction in alcohol preference as it may not have directly affected alcohol consumption.

In Thomsen et al. (2018)’s study, liraglutide was observed to significantly reduce alcohol consumption compared to the vehicle group (F (1,22) = 17.3, p = 0.0004). This intake was related to day (F (11,242) = 5.08, p < 0.0001), and a treatment by day interaction (F (11,242) = 2.44, p = 0.007), indicating that the treatment had a sustained impact on alcohol consumption until the second day, which is in accordance with Marty et al (2020)’s findings. Following the second day, a distinguishable rebound effect was briefly observed before alcohol intake resumed to vehicle-levels and the liraglutide treatment effect disappeared.

Vallof et al. (2015) focused primarily on the effects of liraglutide on alcohol-associated conditioned place preference (CPP) and alcohol-induced dopamine release. Liraglutide treatment at a dosage of 0.1mg/kg significantly reduced alcohol-induced CPP and dopamine release (P = 0.0065). Liraglutide was also found to significantly reduce both alcohol intake (P < 0.01) and alcohol preference (P < 0.05), compared to the vehicle group. In high-alcohol consuming rats, liraglutide was found to significantly reduce alcohol consumption (P = 0.0056), without affecting water and total fluid intake, or alcohol preference. Conversely, no significant effect on alcohol intake (P = 0.1100) and alcohol preference (P = 0.5405) was observed in low-alcohol consuming rats, relative to the vehicle treatment. In a relapse-like drinking model, marked deprivation effects associated with liraglutide’s effects on alcohol intake and preference were detected in control group rats and absent in the treatment group after 24 hours. While the reductions in alcohol intake appeared transient over time, the effects were prolonged with longer treatment periods. Significant reductions were prominent in the earlier sessions but soon returned to baseline levels in later days, especially with the 0.05mg/kg dosage. The 0.1mg/kg group experienced prolonged effect responses for 2-3 days before returning to vehicle-like levels.

We identified two clinical studies evaluating the effects of GLP-1RAs (e.g., semaglutide and on alcohol consumption in patients diagnosed with AUD (Klausen et al., 2022; Quddos et al., 2023). Results from the clinical literature indicate that there are mixed conclusions drawn on the role of GLP1-RAs on alcohol-related behaviours (e.g., alcohol consumption).

In a randomised control trial by Quddos et al. (2023) (n = 153), semaglutide administration was associated with significant attenuation of alcohol consumption in comparison to control and non-treatment groups (p < 0.001). Notably, binge-drinking behaviours were reduced following treatment, as supported by the binomial model [i.e., Semaglutide vs. Control (B = −2.0517, SE = 0.6002, p < 0.001)]. Additionally, drinks per episode of regular usage were significantly reduced after participants initiated medication consumption (B = 2.3443, SE = 0.2668, p < 0.001). Quddos et al. (2023) further identified significant reduction in AUDIT scores within (i.e., pre- and post- starting medications) and between (i.e., medicated vs control) groups following semaglutide administration.

In an intervention study, Klausen et al. (2022) reported that weekly exenatide administration did not significantly reduce heavy alcohol consumption days relative to the placebo group. Notwithstanding, an exploratory BMI subgroup analysis indicates that exenatide significantly decreased heavy drinking days by 23.6% (95% CI [−44.4, −2.7], P = 0.034) and reduced cumulative alcohol consumption by 1,205g within a month (95% CI [−2,206, −204], P = 0.026) in obese patients (BMI>30 kg/m², n = 30) relative to placebo group. However, these trends were not observed in patients with a BMI<25 kg/m² (n = 52). Instead, exenatide administration increased heavy drinking days by 27.5% (95% CI [4.7, 50.2], P = 0.024) relative to placebo, though fMRI Alcohol Cue Reactivity analysis revealed a significant reduction in cue reactivity in ventral striatum, dorsal striatum, and putamen after 26 weeks. Whole-brain analysis further revealed initial cue reactivity differentiation in certain regions compared to healthy controls, but these differences were resolved by week 26, with exenatide additionally reducing activation in the temporal lobe, hippocampus, and parahippocampus. Nonetheless, there is not adequate information on the effects of GLP-1RAs on alcohol consumption and associated behaviours.