Efficacy of Varenicline in the Treatment of Alcohol Dependence: An Updated Meta-Analysis and Meta-Regression

Background: Although varenicline has been used for alcohol dependence (AD) treatment, its efficacy for this condition remains controversial. Aims: This systematic review and meta-analysis of randomized controlled trials (RCTs) assesses the efficacy and safety of varenicline in patients with AD. Methods: PubMed, Cochrane Library, ScienceDirect, Web of Science, and ThaiLis were systematically searched. RCTs investigating the efficacy and safety of varenicline in patients with AD were included. Study selection, data extraction, and quality assessment were independently performed by two authors. The Jadad score and Cochrane risk of bias were used to assess the quality of the included studies. Heterogeneity was assessed using I2 and chi-squared tests. Results: Twenty-two high-quality RCTs on 1421 participants were included. Varenicline significantly reduced alcohol-related outcomes compared with placebo based on percentage of abstinent days (standardized mean difference [SMD] 4.20 days; 95% confidence interval [CI]: 0.21, 8.19; p = 0.04), drinks per day (SMD −0.23 drinks; 95% CI: −0.43, −0.04; p = 0.02), drinks per drinking day (SMD −0.24 drinks; 95% CI: −0.44, −0.05; p = 0.01), craving assessed using the Penn alcohol craving scale (SMD −0.35; 95% CI: −0.59, −0.12; p = 0.003), and craving assessed using the alcohol urge questionnaire (SMD −1.41; 95% CI: −2.12, −0.71; p < 0.0001). However, there were no significant effects on abstinence rate, percentage of drinking days, percentage of heavy drinking days, alcohol intoxication, or drug compliance. Serious side effects were not observed in the varenicline or placebo groups. Conclusion: Our results indicated that AD patients treated with varenicline showed improvement in percentage of very heavy drinking days, percentage of abstinent days, drinks per day, drinks per drinking day, and craving. However, well-designed RCTs with a large sample size and long duration on varenicline treatment in AD remain warranted to confirm our findings.


Introduction
Excessive alcohol consumption remains a crucial public health concern that affects both the economy and society. There are approximately 3 million alcohol-related mortalities annually worldwide, accounting for 5.3% of all mortalities, and the mortality continues to increase, especially among those aged 20-39 years [1]. Moreover, alcohol consumption is associated with the occurrence of more than 200 diseases, including psychological disorders and non-communicable diseases, such as cirrhosis, cancer, and cardiovascular diseases [2]. Alcohol dependence (AD) refers to the pattern of continuous and increasing consumption of alcohol often associated with physical and psychological consequences [3]. In the International Classification of Diseases (ICD)-11, alcohol use disorder (AUD) is either diagnosed as "alcohol dependence" or a "harmful pattern of the use of alcohol." Moreover, the ICD-11 expanded on a group of substances, such as alcohol, nicotine, caffeine, and Int. J. Environ. Res. Public Health 2023, 20, 4091 4 of 25 treatment by assessing abstinence rate, percentage of drinking days, percentage of heavy drinking days, percentage of very heavy drinking days, percentage of abstinent days, drinks per day, drinks per drinking day, and alcohol craving using questionnaires, such as the Penn alcohol craving scale (PACS) and the alcohol urge questionnaire (AUQ). The secondary outcome was varenicline safety, which included AEs. The outcomes reported by the dichotomous scale were estimated using the relative risk (RR) with a 95% confidence interval (CI). Moreover, the standardized mean difference (SMD) was used to estimate the treatment effects for continuous parameters. We utilized SMD since the included studies had differences in the baseline characteristics of participants and varying durations.
Two statistical models were used for the analysis of the results: the fixed-effects model and random-effects model. The former was used when there was no significant difference among the studies included in the MA, and the latter, specifically the DerSimonian and Laird random-effects model, was used when there was a significant level of heterogeneity between the studies. The level of heterogeneity was estimated using the I 2 value, where I 2 < 50% indicated low heterogeneity and I 2 = 50% or higher indicated high heterogeneity [23].
Publication bias was assessed using Eager's weighted-regression statistics and visual inspection of funnel plots [24,25]. The DerSimonian and Laird random-effects model [26] was employed for all analyses. Statistical analysis was performed using Stata software version 14 (StataCorp, College Station, TX, USA) and Review Manager (RevMan) version 5.3.5.
Sensitivity analysis was performed by changing the effect model to ensure robustness of the results [27]. In addition, we conducted subgroup analyses based on four factors, namely duration of treatment, dose of varenicline, levels of alcohol addiction before enrollment, and AD participants alone versus those who are smokers.
Meta-regression analysis was conducted to evaluate the associations between the effect size and potential modifier variables, including dose and duration of varenicline treatment. We also performed a weighted fixed-effect meta-regression analysis using the unrestricted maximum likelihood model.

Study Characteristics
The total number of participants among the included studies was 1421 (720 in the varenicline group and 701 in the placebo group). The number of participants in the included studies ranged from 10 to 200. Moreover, all included studies enrolled participants aged ≥18 years, with a mean age of 38.55 ± 6.40 years. The treatment period ranged between 8 and 112 days. Eighteen studies were performed on AUD patients, whereas four studies were on AUD with smoking patients. The dose of varenicline ranged from 1 to 2 mg/day. We included previous studies that investigated the effect of varenicline on AD in all patient groups, among which two studies were conducted on heavy alcohol drinkers with depression and 35 were on patients with schizophrenia who were both alcoholics and smokers [35]. Other characteristics of the included studies are presented in Appendix B.

ROB in Included Trials
The ROB assessment is shown in Figure 2. All studies that clarified the randomization were described as RCTs. Most of the described methods were of random sequence gener-  The methodological quality of the studies was generally high, with the Jadad score ranging from 3 to 5. Only one study by Meszaros et al. (2013) [35] was given a score of two since it was not defined as a double-blind study and the randomization process was not appropriately described.

Primary Outcomes Abstinence Rate
The pooled results of two RCTs [31,41], including 245 patients, showed no significant difference in the abstinence rate between varenicline and placebo (RR = 0.70; 95% CI: 0. 21-2.35). In addition, the result of the test for heterogeneity between the studies was not significant (p = 0.59; I 2 = 0.0%) ( Table 1). Table 1. Outcome comparison between varenicline and placebo.

Percentage of Drinking Days
The aggregated results of two RCTs that included 66 patients with AD showed no significant difference in the percentage of drinking days between varenicline and placebo (SMD: −0.10; 95% CI: −0.58, 0.38). The test for heterogeneity was not significant (p = 0.90; I 2 = 0.0%) ( Table 1).

Alcohol Craving Evaluation
Alcohol craving was evaluated using four different types of questionnaires: the obsessive-compulsive drinking scale (OCDS), PACS, AUQ, and the visual analog scale (VAS). The scores of the questionnaires ranged from 0 to 100. Six studies used the OCDS questionnaire and the pooled analysis showed no significant difference in alcohol craving between the varenicline and placebo groups (n = 341; SMD: −0.25; 95% CI: −0.72, 0.22; p = 0.22; I 2 = 73%). The VAS showed that the varenicline-treated group had decreased alcohol craving but the difference was not statistically significant compared to the placebo group (n = 182; SMD: −0.26; 95% CI: −0.55, 0.04; p = 0.09; I 2 = 0.0%).

Sensitivity Analysis
In this study, the sensitivity analysis was performed by changing the effect model to establish the sensitivity of each outcome. The results were similar to those of the main analysis; this confirmed that the main results were robust and reliable.

Subgroup Analysis
We also conducted a subgroup analysis to reinforce the results of the main MA. This analysis was divided into four categories: (1) varenicline dose (1 and 2 mg per day), (2) duration of treatment (<30 days, 30-90 days, and >90 days), (3) alcohol consumption level (very high, high, medium, low, and no daily alcohol intake), and (4) participant characteristics (AD alone vs. AD with smoking).  The results showed that the low dose (1 mg/day) of varenicline significantly decreased alcohol intoxication but there was no significant difference in alcohol craving as evaluated by AUQ scores compared with that of the placebo group. The treatment with varenicline for 30-90 days did not improve alcohol craving as evaluated by VAS scores but there was a significant improvement in alcohol craving as evaluated by PACS scores for treatment duration >90 days. Additionally, the durations <30 days, 30-90 days, and >90 days resulted in non-significant decreases in percentage of abstinent days, drinks per day, and drinks per drinking day. The subgroup analysis of alcohol consumption levels indicated that varenicline had significant effects on reducing heavy drinking days, alcohol intoxication, and drug compliance for participants with low levels of alcohol consumption.
The subgroup analysis based on participant characteristics showed that varenicline treatment significantly improved alcohol craving scores as evaluated by PACS, AUQ, and VAS, as well as alcohol intoxication and drug compliance outcomes for participants with both AD alone and AD with smoking. The overall and stratified analysis results are presented in Appendix C.

Meta-Regression
Meta-regression was used to evaluate the association between primary outcomes and the duration of varenicline administration. The results from the random-effect metaregression showed three significant associations between duration of varenicline use and

Discussion
Varenicline has been shown to have potential benefits in reducing alcohol consumption in patients with AD. Our updated SR and MA aimed to summarize the available clinical evidence regarding the efficacy and safety of varenicline for AD treatment. The results indicated that varenicline had a positive impact on measures, such as percentage of abstinent days, drinks per day, drinks per drinking day, and alcohol craving. Although the treatment was well tolerated, the incidence of serious AEs was not reported. Meta-regression analysis suggested an association between varenicline dose and outcomes, such as percentage of heavy drinking days, AUQ score, and alcohol intoxication. However, it is important to note that our study included a larger number of trials and participants than did previous SRs [12,13]. Further well-designed RCTs with larger sample sizes and longer treatment periods are needed to confirm the results and assess the overall safety and efficacy of varenicline for AD treatment.
The mechanism of lowering alcohol cravings remains unclear but may be related to the nicotinic receptor. Davis et al. (2006) reported that alcohol acts on nicotine receptors, resulting in the alcohol craving and drinking behavior observed in both alcoholics and smokers. Varenicline had inhibitory effects on nACH, possibly reducing cravings for both cigarettes and alcohol [35,38].
Our results are consistent with those of Oon-Arom (2019) [12], who analyzed nine RCTs and showed that varenicline reduced the percentage of heavy drinking days compared to placebo but this difference was not significant. However, varenicline significantly improved the consumption of drinks per day. Results from this MA demonstrated that varenicline significantly reduced the consumption of drinks per drinking day. The safety outcomes of the previous study were not assessed, however, the acceptability test using the dropout rates method was conducted. The findings showed that the main reason for requesting to terminate the study was adverse reactions from varenicline use. In contrast to this study, there was no significant difference in the AE of treatment with varenicline in any dosage range or duration compared to placebo.
Meanwhile, the comparison with the MA of Gandhi et al., (2020) [13] that included 10 studies showed that the outcome of percentage of heavy drinking days was significantly reduced with varenicline but did not significantly differ from placebo. The results of drinks per drinking day and the percentage of abstinent days were inconsistent with those of this MA. Varenicline reduced the amount of alcohol consumption on drinking days or the abstinent days, however, there were no differences from the placebo group. Erwin et al., (2014) [42] reported that varenicline affected alcohol outcomes, including decreasing the amount of alcohol consumption and duration of drinking. There are two possible mechanisms to explain the effects of varenicline: (1) varenicline affected the rewarding system, thereby lowering alcohol craving, and (2) varenicline exacerbated the negative effects of alcohol intake, such as headache and dizziness.
Although there were no reports of severe AEs in the varenicline and placebo groups, our MA found that varenicline treatment is accompanied with gastrointestinal effects (nausea/vomiting and abdominal pain) and other AEs associated with CNS (vivid dreams or nightmares).
The possible AE mechanism may be due to varenicline being a partial agonist that activates nACH. Varenicline is a partial agonist of nACH on neurons and stimulates dopamine release, which may affect the chemoreceptor trigger zone, resulting in flatulence, visceral pain, nausea, and vomiting [43,44].
These AEs were found to be comparable to those reported with other medications used for alcoholism treatment, such as naltrexone, acamprosate, and disulfiram. Naltrexone is an opioid receptor antagonist that has been shown to reduce alcohol craving and improve the success of alcohol abstinence. Common side effects of naltrexone include nausea, headache, dizziness, and liver problems. However, it is generally considered safe and well tolerated [45]. Acamprosate is another medication that has been used to treat alcoholism. The most common side effects include diarrhea, headache, and nausea [46]. Disulfiram treatment is associated with unpleasant symptoms, such as flushing, sweating, and headache, after drinking alcohol.
Another study comparing varenicline to naltrexone for the treatment of alcoholism found similar AEs, including nausea, headache, and insomnia. However, the study also found that varenicline was associated with a higher incidence of constipation than naltrexone [47,48]. However, this MA found that the incidence of constipation in the vareniclinetreated group was not different from that in the placebo group. It is important to note that the AEs of varenicline may vary depending on the dose and duration of treatment. Additionally, the reported AEs should be considered in light of the potential benefits of varenicline for reducing alcohol craving and consumption.
In addition, these actions may decrease non-rapid eye movement (NREM) sleep. In other words, varenicline may competitively inhibit the binding of acetylcholine receptors. Generally, acetylcholine prevents REM sleep. Therefore, varenicline is associated with the features of REM sleep, with wakefulness as nightmares [44,49].
However, there have been concerns about the safety of varenicline, particularly with regard to the presence of nitrosamines, which are carcinogenic compounds. In recent years, the presence of high levels of nitrosamines in varenicline has led to regulatory actions in several countries, including delays in approvals or restrictions on its use. In 2020, the European Medicines Agency issued a warning about the potential presence of nitrosamines in varenicline and recommended that patients who are taking the medication should continue to do so, as the benefits of quitting smoking outweigh the potential risks from nitrosamines [50]. Moreover, Lang (2023) [51] reported that prescriptions of varenicline were reduced because of this limitation and that the drug was subsequently removed from the market in 2021. The substantial decrease in varenicline use after the drug's recall represents potential lost opportunities for nicotine cessation with likely immediate and long-term adverse health outcomes. This may further affect varenicline use for AD.
The results of the meta-regression analysis herein suggest that the duration of varenicline administration is associated with several alcoholic outcomes. Specifically, the results showed significant associations between varenicline use and a decrease in percentage of heavy drinking days, a decrease in AUQ, and a decrease in alcohol intoxication. Previous studies have also explored the effects of varenicline on alcohol consumption. An RCT by Mitchell et al. (2012) [28] found that varenicline was effective in reducing alcohol consumption and craving in heavy-drinking smokers. Another study by McKee et al. (2009) [9] showed that varenicline reduced alcohol craving and the number of drinks consumed per drinking day in heavy-drinking individuals with a history of AD.
However, the results of the current study suggest that varenicline may not have a significant effect on other outcomes, such as percentage of abstinent days, alcohol craving as measured by OCD, alcohol craving as measured by PACS, and drug compliance. This is consistent with previous findings by Gandhi et al., (2020) [13] who found that varenicline had no significant effect on the number of abstinent days from alcohol in heavy-drinking individuals.
The current evidence suggests that varenicline may have potential in reducing certain aspects of alcohol consumption, as demonstrated by the findings in the present study. However, despite these promising findings, there is still a need for more research to fully understand the effects of varenicline on different alcoholic outcomes and to determine the optimal duration of varenicline administration. Moreover, further research is needed to fully understand the effect size and any potential side effects associated with its use. Additionally, there may be other factors, such as comorbidities or individual patient characteristics, that can impact the effectiveness of varenicline as a treatment option.
The strengths of this SR and MA study are as follows: (1) this study is an updated MA that included 22 RCTs, most of which were of high quality and had low ROB; (2) this MA performed a systematic search through five international databases and a Thai database, along with a manual search for unpublished trials; (3) subgroup analysis and meta-regression were performed to determine the effects of variable on outcomes; (4) AEs were pooled by MA, whereas previous SR and MA studies [12,13] did not perform subgroup analysis and meta-regression.
Nevertheless, this study has some limitations: (1) most of the included trials were conducted with a small number of participants and a short-term duration; (2) the studies included various population characteristics and intervention periods; (3) most of the selected RCTs did not report underlying disease and contaminant medication; (4) some outcomes showed publication bias when evaluated using Egger's test and funnel plot. These may be the root cause of the inconsistent results along with the heterogeneity; (5) the abstinence rate and percentage of drinking days were pooled from two RCTs; as only two RCTs may not have enough power to detect significant differences or to make strong conclusions, the results should be interpreted with caution and further research is needed to replicate and validate the findings.

Conclusions
The results from this MA indicated that varenicline is effective for the percentage of abstinent days, drinks per day, drinks per drinking day, alcohol intoxication, and alcohol craving outcomes. Notably, there were no reports of severe effects in both varenicline and placebo-treated groups. The results of this MA suggest that varenicline may have potential as a treatment for AD. However, large-scale, long-term RCTs on the effects of varenicline on AD in combination with other treatments or network meta-analysis to compare treatment efficacy in AD remain warranted.

Conflicts of Interest:
The authors declare no conflict of interest.

Study Protocol
Efficacy of varenicline in the treatment of alcohol dependence: An updated metaanalysis and meta-regression.

Review Question
To perform an updated meta-analysis and meta-regression of the effects of varenicline in patients with alcohol dependence (AD). Moreover, this study assessed the safety of varenicline use in patients with AD.

Searches
The PRISMA guidelines for conducting systematic review were followed. The databases searched were PubMed, Cochrane Library, ScienceDirect, Web of Science, and ThaiLis. The included studies were randomized controlled trials evaluating the effect and/or safety of varenicline in patients with AD.
The search terms were as follows: varenicline, alcohol, ethanol, alcohol use disorder, heavy drinkers, addiction, dependence, abuse, craving, alcoholism, and abstinence.
In addition, a historical search and hand search of references of the included articles were carried out, along with any further material identified for inclusion.
There were no limitations concerning language, place, and time.

Types of studies included
Only randomized controlled trials (RCTs) were included. The studies reported outcomes in terms of abstinence rate, percentage of abstinent days, percentage of drinking days, percentage of heavy drinking days, drinks per day, drinks per drinking day, alcohol intoxication, alcohol craving, and adverse effect.

Condition or domain studied
The meta-analysis and meta-regression investigated the use of varenicline in patients with AD.

Participants/population
Patients with AD.

Context
Primary outcome(s): abstinence rate, percentage of drinking days, percentage of heavy drinking days, percentage of very heavy drinking days, percentage of abstinent days, drinks per day, drinks per drinking day, and alcohol craving using questionnaires, such as the Penn alcohol craving scale (PACS) and the alcohol urge questionnaire (AUQ).

Data extraction (Selection and coding)
From the retrieved abstracts, the principal investigator first considered all titles of articles and selected RCTs evaluating the efficacy of varenicline in AD. The abstracts of all selected articles were assessed using an abstract evaluation form by two authors working independently. Following this, the two authors independently evaluated the full details of all selected articles using a data extraction form. The extracted data included the following items: authorship, year of publication, location/region of study, population, type of economic analysis, strategies assessed (intervention versus comparator), abstinence rate, percentage of drinking days, percentage of heavy drinking days, percentage of very heavy drinking days, percentage of abstinent days, drinks per day, drinks per drinking day, and alcohol craving using questionnaires, such as the Penn alcohol craving scale (PACS) and the alcohol urge questionnaire (AUQ), and adverse events. A third author's opinion was sought if disagreements occurred between the two researchers.

Risk of bias (quality) assessment
The quality of the included studies assessed using the Jadad's scale was used as a guideline to evaluate methodological quality of included studies. Risk of bias in individual studies was assessed using the risk of bias tool of Cochrane Handbook for Systematic Reviews of Interventions. The approval of the two authors for the selected extracted data was sought. The opinion of a third author was requested if a disagreement occurred between the two researchers.
Strategy for data synthesis Data were extracted by two authors working independently and added into a table for analysis. Efficacy between the varenicline usage and the placebo groups was statistically tested using relative risk (RR) with a 95% confidence interval (CI) for the outcomes reported by the dichotomous scale. Moreover, the standardized mean difference (SMD) was used to estimate the treatment effects for continuous parameters. The heterogeneity of included studies was examined using the Q-statistic, while results were analyzed using the I-squared statistic. Random-effects modelling was used if the included studies were heterogeneous and fixed-effects modelling was used if homogeneity was found. The publication bias was examined via Eager's weighted regression statistics and the means of funnel plot asymmetry.

Analysis of subgroups or subsets
Subgroup analysis was performed based on four factors, namely duration of treatment, dose of varenicline, alcohol consumption levels before enrollment, and AD participants alone versus those who are smokers.

Meta-regression
Meta-regression was employed to evaluate associations between the effect size and potential modifier variables, which included dose and duration of varenicline treatment. We performed a weighted fixed-effect meta-regression using the unrestricted maximum likelihood model.   Ad-libitum alcohol consumption, alcohol craving before drinking (tonic craving), and alcohol craving after drinking (AUQ scale)            Remark:  = yes,  = no.   Ad-libitum alcohol consumption, alcohol craving before drinking (tonic craving), and alcohol craving after drinking (AUQ scale) Week days of alcohol use, presence/absence of alcohol use, numbers of heavy drinking days presence/absence of heavy drinking, alcohol craving (PACS scale), and addiction severity index (ASI)        Ad-libitum alcohol consumption, alcohol craving before drinking (tonic craving), and alcohol craving after drinking (AUQ scale)