The main finding of the present study is that participants treated for smoking cessation drink significantly less alcohol after 12 weeks of treatment with dulaglutide compared with placebo, without correlation to smoking status. Our results thus support the hypothesized effect of GLP-1 agonists on alcohol intake in humans.

The role of GLP-1 in reward-related processes has been demonstrated in preclinical studies. GLP-1 receptors are present in the mesolimbic reward system (6) and their activation attenuates alcohol-induced reward and reduces alcohol intake in animals (7, 8, 11, 25). Translating these findings into humans has proven difficult. Treatment with exenatide for 26 weeks compared to placebo did not lead to a greater reduction in heavy drinking days in a randomized controlled trial of patients with AUD. It did, however, decrease fMRI cue reactivity and central dopamine transporter availability (17); hence, there is evidence that GLP-1 agonism affects the human reward system. A potent placebo response, which is often the case in trials for addiction treatment (26), might have reduced the effect size in the above-mentioned study (17) as well as in our main trial (21). The strong placebo response in the study by Kruse-Klausen et al. may in part be explained by the fact that all participants received standard AUD behavioral treatment sessions every second week throughout the trial (17). Indeed, in the main analysis of the present study, participants in both groups achieved unusually high abstinence rates from cigarettes (63% in the dulaglutide group, 65% in the placebo group), independent of treatment allocation. The fact that we found a significant group difference in alcohol intake at week 12 might be explained by participants’ focus on smoking cessation rather than on reducing their drinking, and participants did not suffer from AUD as a primary disorder. This might also explain why results from animal studies cannot be extrapolated to humans. Animals, unlike participants in clinical trials, have no intrinsic motivation to reduce consumption of alcohol and thus display no placebo effect.

The population of the present study was predominantly obese (BMI > 29.9 kg/m2; 90.8% overall, 87.0% in the placebo group, 95.2% in the dulaglutide group). Interestingly, although Kruse-Klausen et al. found no overall group difference regarding heavy drinking days in AUD patients in the above-mentioned randomized controlled trial, subgroup analyses revealed that exenatide significantly reduced heavy drinking days in obese individuals (defined as BMI > 30 kg/m2) (17). Furthermore, exenatide inhibited brain responses to food cues in obese individuals, but not in lean individuals in a cross-sectional study (27). GLP-1 agonists might therefore affect the reward system differently in obese patients compared with lean individuals.

Participants in our study received standard smoking cessation therapy, i.e., behavioral counselling and varenicline. Behavioral counselling is a form of psychotherapy implemented in treatment of SUD, including AUD. Varenicline has been found to reduce alcohol intake, especially in smokers, but also independently of smoking status (28). These concomitant therapies might have affected alcohol intake in our population. However, all participants received varenicline and behavioral counselling regardless of treatment allocation, and thus we consider group differences in alcohol consumption to be due to the effect of treatment allocation.

In pursuit of alternative reinforcers, patients quitting smoking may increase their consumption of alcohol (29). Furthermore, alcohol may lessen withdrawal symptoms from nicotine (30); hence, one could hypothesize that participants who try to quit smoking would not only drink more alcohol for reward effects, but also curb withdrawal symptoms. Preclinical data suggest that GLP-1 can attenuate withdrawal, as experiments with GLP-1 agonists reduced withdrawal-induced anxiety in rodents (31, 32). If the same were true in humans, GLP-1 agonists would diminish the urge to consume alcohol when experiencing withdrawal symptoms. In the present study, there was no increase in alcohol consumption in the placebo group. One possible explanation might be that participants in both groups received a concomitant therapy with varenicline, which, as mentioned above, has also been reported to reduce alcohol consumption (28, 33). In type 2 diabetes and obesity, GLP-1 agonists are prescribed due to their effect on appetite and homeostatic feeding, and clinical studies report a reduction in fluid intake (34, 35). We did not assess calorie or fluid intake and thus cannot rule out that these mechanisms had an effect on our results. However, 2 facts lead us to believe that the reduction in alcohol consumption under dulaglutide is not solely due to a reduction in overall energy and fluid intake. First, in rodents, GLP-1 agonists reduce consumption of substances that have no caloric value, i.e., psychostimulants; and second, different preclinical studies have shown a decrease in alcohol intake independent of fluid or calorie intake (8, 25, 36). Furthermore, changes in consumption of alcohol and weight changes were not correlated (see supplemental material).

Gastrointestinal symptoms are common side effects of GLP-1 agonists (37–40) and might have affected alcohol intake. However, as expected, these symptoms were only transient and more common in the dulaglutide group at week 2, but comparable at week 12, when alcohol consumption was assessed (21).

Smoking and drinking are closely associated (41–43). It has been hypothesized that nicotine deprivation leads to a higher intake of alcohol in pursuit of alternative reinforcers (29) and attenuates withdrawal symptoms (44). Accordingly, acute nicotine deprivation has been reported to increase the urge to consume alcohol and actual alcohol intake (30). However, other studies indicate that smoking cessation does not lead to increased alcohol intake (45–47); some data even suggest that reducing nicotine consumption may have beneficial effects on alcohol consumption (48). Our finding that smoking status at week 12 was not correlated with changes in alcohol consumption is in line with the majority of literature. However, in contrast to the studies reviewed, most of our participants were not heavy drinkers or AUD patients. Furthermore, high abstinence rates from nicotine in the placebo group might have masked a correlation. Alcohol is a well-established smoking trigger (49, 50). Heavy and binge drinkers are less likely to quit smoking, especially in the short term (51–54). The influence of moderate alcohol consumption on smoking cessation success remains elusive; some studies found no difference in quit rates among moderate and nondrinkers (41, 53–55). In other trials though, and in accordance with our study, moderate drinkers displayed higher abstinence rates compared with those who did not drink (52, 56). Nondrinkers were found to have fewer smokers in their social network (52). Thus, it has been hypothesized that those who start smoking despite their environment not actively enabling them are those who find smoking more rewarding (51). Indeed, smoking patterns of nondrinkers have been described as “heavy, automatic, and characterized by a sense of loss of control” (52).

In analogy to the effects on alcohol-mediated reward, in preclinical studies, GLP-1 agonism has been found to also attenuate reward induced by other substances, i.e., psychostimulants (12–14, 57), nicotine (15, 16), and to some extent opiates, though data on the latter are contradictory (58–60). An experimental study testing the effects of exenatide on cocaine consumption in 13 non–treatment-seeking individuals suffering from cocaine use disorder failed to show a reduction in cocaine self-administration and found no changes in subjective cocaine-related effects (19). However, cocaine consumption has been found to decrease GLP-1 serum levels in humans, and the authors hypothesized that this mechanism leads to sustained cocaine consumption (19, 20). To elucidate the exact mechanisms, and to investigate whether GLP-1 plays a role in addiction to other drugs of abuse, such as cannabis or opioids, more data are needed. In our analysis we did not find an effect of dulaglutide on the consumption of drugs, probably because the number of participants consuming drugs (n = 30) was too small.

The present study has limitations. First, our participants did not per se suffer from AUD or SUD and the subgroup of heavy drinkers was too small to provide valuable results. Whether our findings can be extended to heavy- or binge-drinking smokers or nonsmoking AUD patients, and how smoking status influences the effect of GLP-1 agonists on alcohol intake needs to be investigated. Second, our primary endpoint of alcohol consumption is self-reported and therefore subjective. However, we were not able to assess objective measurements to quantify alcohol intake, such as biomarkers, in the present study. Furthermore, we did not assess drinking patterns, such as percent heavy-drinking days, abstinence, time spent drinking, or the context in which our participants drank. Third, in our study, participants received dulaglutide for only 12 weeks. Preclinical data show that alcohol intake was reduced for 3 weeks after discontinuation of dulaglutide in male but not female rats (9). Whether termination of treatment leads to a rebound effect in humans or whether the effects are sustained, and whether sex plays a role in posttreatment effects will need to be explored in future studies.

In conclusion, our analysis showed an effect of a GLP-1 agonist on alcohol intake in humans and our results thus strengthen the rationale for implementing GLP-1 agonists in pharmacological AUD treatment. However, it has to be considered that AUD is characterized by a wide range of drinking behaviors and prevalent comorbid psychiatric disorders, including substance abuse. Thus, treatment strategies need to be multidimensional and personalized, taking into account psychosocial components. To identify the GLP-1 agonist’s role in the treatment of AUD and SUD, further studies are needed, elucidating which patients would benefit and whether GLP-1 agonists improve long-term abstinence.