Pharmacogenetics of alcohol use disorder treatments: an update

Abstract

Introduction

Alcohol use disorder (AUD) is a highly prevalent; costly economically, socially, and interpersonally; and grossly undertreated disorder. The low rate of utilization of pharmacological treatments with demonstrated efficacy is particularly noteworthy. This is due, in part, to the modest efficacy of the medications approved to treat the disorder. One approach to increasing the utility and safety of these medications is to use precision medicine, which seeks to identify patients for whom specific medications are likely to have the most robust therapeutic effects and the fewest adverse effects.

Areas Covered

Here we review the current literature on the pharmacogenetics of AUD treatment. We cover both laboratory studies and clinical trials that have provided valuable insights into the mechanisms and value of precision-based care for AUD. We discuss studies of genetic moderators for personalizing pharmacotherapy with medications approved by regulatory agencies in the United States and Europe and those that are used off-label to treat AUD.

Expert Opinion

Pharmacotherapy can be a useful component of AUD treatment. Currently, the evidence regarding genetic predictors of medication efficacy is very limited. Thus, precision medicine is not yet ready for widespread clinical implementation. Further research is needed to identify candidate genetic variants that moderate the response to both established and novel medications in development for this goal to be achieved. The growing availability of large-scale, longitudinal datasets that enable the synthesis of genetic and electronic health record data could provide important opportunities to develop this area of research.

1. Introduction

Alcohol use disorder (AUD) is highly prevalent, with estimates ranging from 5.6% [1] to nearly 14% [2] of U.S. adults meeting criteria for the disorder during the preceding year and approximately 6% of the population endorsing heavy drinking in the past month [3]. The fifth and latest edition of the Diagnostic and Statistical Manual (DSM), published in 2013 by the American Psychiatric Association, changed the previously used distinct diagnostic categories of alcohol dependence and abuse into a unidimensional AUD diagnosis with a severity specifier. The first and second editions of the DSM, first published in 1952, used the diagnostic term “alcoholism,” however, as this term is imprecise and ambiguous, the term AUD is preferred. To be diagnosed with AUD, an individual must report problematic alcohol use leading to significant distress or impairment, endorsing at least two of the 11 diagnostic criteria [4], which are grouped into impaired control, social impairment, risky use, and pharmacological categories. A severity specifier based on the number of criteria endorsed divides AUD into “mild” (2–3 criteria), “moderate” (4–5 criteria), and “severe” (6 or more criteria).

AUD, which usually has its onset before age 40 [2], is commonly associated with high rates of psychological and physical comorbidities, social and familial disruption, and decreased work productivity. Although the central feature of AUD is impaired control over drinking and it is characterized by episodic heavy drinking and/or the daily consumption of large amounts of alcohol, it is highly heterogeneous in its presentation and course.

The estimated heritability of AUD ranges from 40–70% [5, 6, 7], which also reflects a substantial environmental contribution to risk of the disorder [8]. Various endophenotypes, including the subjective response to alcohol and neurobiological vulnerabilities, which are partially heritable, can be strong predictors of who will develop an AUD [9]. Understanding the factors contributing to the development and the pathophysiology of AUD can inform treatments for the disorder [10].

Alcohol is a unique substance of abuse because it does not have a specific receptor on which it acts, exerting its central nervous system effects through interactions with a host of neurotransmitter systems. Research has examined the impact of alcohol on opioidergic, dopaminergic, GABAergic, glutamatergic, and serotonergic neurotransmission and the genes the encode the proteins in these systems [10, 11]. GABAergic and glutamatergic systems are inhibitory and excitatory, respectively, and dysregulation of these neurotransmitters leads to impaired learning, tolerance, and seizures [11]. The dopaminergic system is involved in motivation, reinforcement, motor control, and hormonal response. The ingestion of alcohol increases dopaminergic transmission [11]. Alcohol administration also increases endogenous opioid activity, which regulates mood, pain, emotions, and reinforcement [11]. Finally, serotonin is associated with mood, thought, and sleep and eating behaviors, which are modified by its release in response to small amounts of alcohol [11]. The actions of these various neurotransmitter systems are intertwined, further complicating our understanding of the impact of alcohol on the brain and its pharmacogenetic effects. For example, the release of serotonin stimulated by alcohol consumption may in turn increase dopamine release and associated emotional behaviors [12].

Despite AUD’s high prevalence and cost to individuals, families, and society, less than 20 percent of individuals with the disorder seek and receive alcohol treatment [2] and only a small fraction of these individuals are prescribed a medication with demonstrated efficacy in reducing heavy drinking or promoting abstinence [13]. Reasons for this low rate of seeking treatment include its high cost, limited availability, and the inadequate knowledge or skill of providers [14, 15]. As with the course and presentation of AUD, variation in the efficacy of AUD treatment can be influenced by genetic variation. Thus, recently, there has been considerable interest in personalized, stratified, or precision medicine, which is characterized by the tailoring of treatments to individuals based on key characteristics, such as genetics, environment, and lifestyle, to maximize the efficiency and benefit to patients. Of particular interest here are efforts to tailor pharmacotherapy to individuals’ specific genetic profiles in order to enhance treatment adherence (by reducing adverse effects) and improve outcomes (by increasing the therapeutic response).

Currently, there are three medications approved by the U.S. Food and Drug Administration (FDA) for the treatment of AUD: disulfiram, acamprosate, and naltrexone (both oral and injectable formulations). Disulfiram, which was approved in 1949, acts by inhibiting the metabolism of acetaldehyde, a toxic intermediary metabolite of ethanol, causing an unpleasant physiological reaction [16]. Naltrexone, approved by the FDA in 1994, is a μ-, κ-, and δ- opioid receptor antagonist that modulates the dopamine-mediated rewarding effects of alcohol and reduces its consumption [10, 16]. In 2006, a long-acting, injectable formulation of naltrexone was approved for treating AUD. Although the exact mechanism of acamprosate’s effects is unclear, it appears to modulate glutamatergic neurotransmission and was FDA approved in 2004 [17].

Nalmefene and baclofen are approved in the European Union and France, respectively, but not in the United States. Nalmefene, a μ- and δ-opioid receptor antagonist and a κ-opioid receptor partial agonist, was efficacious in reducing binge drinking and overall alcohol consumption in three European trials [18, 19, 20]. Baclofen is a GABA_B receptor agonist that has shown efficacy in maintaining abstinence [16, 21]. At present, there are also over 30 compounds in Phase I and II clinical trials in the United States, 75% of which are FDA approved for other indications. The preclinical and clinical safety evaluations conducted in support of the approval of these medications will reduce the time and expense of developing them for use in treating AUD [22].

Strategies to enhance treatment response to the various medications under investigation include personalized treatment through the application of pharmacogenetics, which focuses on the genetic moderation of the pharmacokinetic and pharmacodynamic effects of a medication. Pharmacodynamic effects refer to the drug’s effect on the body, which include receptor-mediated effects that produce both therapeutic and adverse effects. Pharmacokinetic effects refer to the impact of the body on the drug and include effects on its absorption, distribution, metabolism, and excretion. Here, we review pharmacogenetic studies (published through November 2018), including both randomized controlled trials (RCTs) and human laboratory studies relevant to treating AUD. The review covers both FDA-approved medications and those in development.

2. Pharmacogenetics of AUD Medications

2.1. FDA Approved Medications

2.1.1. Naltrexone

Of the three FDA-approved medications, the pharmacogenetics of naltrexone has been studied most extensively. An opioid receptor antagonist, naltrexone exerts a small, but statistically reliable, effect in treating AUD [23, 24], principally by reducing heavy drinking and craving [25]. However, not all studies have shown naltrexone to be efficacious in treating AUD (e.g., [26]). In an effort to personalize AUD treatment and enhance naltrexone’s efficacy, pharmacogenetic studies of the medication (summarized in Table 1) have examined the moderating effects of the Asn40Asp single nucleotide polymorphism (SNP) in exon 1 of the μ-opioid receptor gene (OPRM1). OPRM1 encodes a protein in the endogenous opioid system that is thought to be involved in pleasure and reward pathways. It has been hypothesized that variation in OPRM1 contributes to the risk of substance use disorders and to food preferences, pain perception, and physiological stress reactivity [27, 28]. The first report suggesting such an effect came from a secondary analysis of 141 participants from three 12-week, placebo-controlled RCTs [29]. The analysis showed that European-American carriers of the minor Asp40-allele were associated with a higher rate of relapse to heavy drinking when treated with placebo than naltrexone [odds ratio (OR)=3.52], with no difference in relapse rate in individuals homozygous for the Asn40 allele [29]. Asp40 carriers treated with naltrexone also took longer to relapse to heavy drinking than Asn40 homozygotes treated with naltrexone (OR=2.79). However, it should be noted that, in this analysis, the interaction of gene by medication was not significant for any treatment outcome measure. Conversely, such an interaction was seen in the COMBINE study [30], where a secondary analysis was conducted in 604 Caucasian participants who provided genetic samples [31]. The analysis demonstrated a strong moderating effect of the OPRM1 genotype, such that Asp40-allele carriers treated with naltrexone were significantly more likely to show a good clinical outcome than either Asn40-allele homozygotes or either of the genotype groups treated with placebo. Asp40-allele carriers treated with naltrexone also reported a significantly lower percentage of heavy drinking days and a significantly higher percentage of abstinent days than either Asn40 homozygotes or the two placebo groups.

Table 1.

Pharmacogenetic studies of naltrexone for treating alcohol use disorder.

Reference	SNP(s)	N	Population Group	Diagnosis	Effect
Oslin et al. (2003)1	rs1799971	141	European Americans	Alcohol dependence	Asp40 carriers treated with naltrexone had a lower rate and a longer time to relapse to heavy drinking.
Anton et al. (2008)2	rs1799971	604	European Americans	Alcohol dependence	Asp40 carriers treated with naltrexone had a lower percentage of heavy drinking days and greater likelihood of “good clinical outcome.”
Gelernter et al. (2007)2	rs1799971 and 5 other opioid receptor SNPs	215	European and African Americans	Alcohol dependence	No SNP x medication interaction for any of the 6 SNPs examined.
Kim et al. (2009)3	rs1799971	63	Korean	Alcohol dependence	Asp40 carriers had a lower rate and a longer time to relapse to heavy drinking.
Ooteman et al. (2009)4	rs1799971; SNPs in DRD2, GABRA6, and GABRA2	108	Dutch	Alcohol dependence	No gene x medication effect
Coller et al. (2011)3	rs1799971	100	Australian	Alcohol dependence	No gene x time effect
Arias et al. (2014)5	rs1799971; rs1611115 in DBH	107	European Americans	Alcohol dependence with comorbid Axis I disorder	No gene x medication interaction. T-allele carriers of DBH reported less heavy drinking when treated with naltrexone.
Foulds et al. (2015)3	rs1799971	108	European and Maori descent	Alcohol dependence with major depression	No gene x medication interaction
Oslin et al. (2015)6	rs1799971	221	Predominately European Americans	Alcohol dependence	No gene x medication interaction
Schacht et al. (2017)6	rs1799971	152	European Americans and Asian Americans	Alcohol dependence	No gene x medication interaction. Asp40 carriers treated with naltrexone had a more rapid return to heavy drinking after treatment ended.

¹Secondary analysis of 3 separate trials

²Secondary analysis of a larger, placebo-controlled trial

³Open-label, no control group

⁴Open-label comparison of naltrexone and acamprosate

⁵Secondary analysis of randomized trial comparing naltrexone, placebo, disulfiram or disulfiram with naltrexone

⁶Placebo-controlled, with randomization stratified on genotype

Despite these promising initial findings, subsequent RCTs that examined the moderating effect of the Asn40Asp SNP on the response to naltrexone have yielded conflicting findings. One example of this involves a secondary analysis of data from the Department of Veterans Affairs Cooperative Study 425, which examined the efficacy of 12 weeks or one year of naltrexone treatment in male veterans with alcohol dependence [26]. In a subset of study participants (N = 220 or 35.1% of the total sample), seven polymorphisms in OPRM1, OPRD1, and OPRK1 were examined as potential moderators of naltrexone response [32]. None of the SNPs were found to moderate the medication response.

Although clinical trials typically seek to limit participant heterogeneity by excluding individuals with co-occurring psychiatric disorders, two trials tested pharmacogenetic moderation of naltrexone’s effects in dually diagnosed individuals. Arias et al. [33] recruited 107 European-American, alcohol-dependent, male veterans with a comorbid Axis I disorder to participate in a 12-week trial in which they were randomized to receive treatment with naltrexone, placebo, disulfiram and naltrexone, or disulfiram and placebo. Study participants were genotyped for the Asn40Asp SNP in OPRM1 and rs1611115 in the dopamine β-hydroxylase (DBH) gene, which encodes a key enzyme in the conversion of dopamine to norepinephrine. Inhibition of DBH is associated with psychosis and major depression [34]. Despite there being no effects of the OPRM1 SNP, the study showed that the DBH variant moderated the response to naltrexone, such that T-allele carriers who received naltrexone were less likely to report heavy drinking. In a second study, 108 alcohol-dependent patients with major depression were treated with naltrexone and randomly assigned to receive citalopram or placebo [35]. The results showed no moderating effects of the Asn40Asp SNP on any of the alcohol consumption outcome measures tested, including percent days abstinent (Cohen’s d=0.06), drinks per drinking day (d=0.00), and heavy drinking days (d=0.04).

Because of the bias inherent in secondary analyses, two RCTs prospectively genotyped individuals for the Asn40Asp SNP, stratifying the randomization to naltrexone or placebo on genotype. Both trials oversampled individuals with an Asp40 allele. The first study, a 12-week trial in 221 alcohol-dependent subjects from five sites across Pennsylvania, showed no evidence of a gene-by-medication interaction on heavy drinking, the primary outcome (OR=1.10 for Asp40 carriers treated with naltrexone versus placebo), abstinence, or any other alcohol-related outcome, including craving [36]. The second study, a 16-week trial in 151 alcohol-dependent individuals, showed an overall reduction in heavy drinking in the naltrexone group, with no evidence of moderation by OPRM1 genotype [37]. The authors of this report noted that, although the genotype by medication interaction was not significant, the effect size (Cohen’s d) for the difference between naltrexone and placebo was 1.1 for Asp40-allele carriers, but only 0.19 for Asn40-allele homozygotes. In addition, there was a significant interaction of genotype, medication, and time, with Asp40-allele carriers treated with naltrexone showing an increase in the rate of heavy drinking once the medication was stopped, whereas heavy drinking in the other groups remained stable [37]. In an open-label study, 100 Australian alcohol-dependent individuals treated with naltrexone for 12 weeks significantly decreased both self-reported and objective indicators of alcohol use and craving from baseline levels. However, there was no evidence of a significant association between the OPRM1 Asn40Asp genotype and any of the outcome measures [38].

Another factor that complicates the interpretation of these findings is the wide variation in the prevalence of the OPRM1 Asp40 allele in different population groups. Individuals of East Asian descent have a much higher prevalence of the allele than other populations (i.e., up to 50% of Japanese compared with approximately 20% in European Americans and 3% in African Americans), thus the potential moderating effects of the SNP could differ substantially by population [39]. In a laboratory study of 29 healthy individuals (59% Asian ancestry), naloxone was administered intravenously to investigate OPRM1 mediation of HPA-axis activation [40]. Results showed that European-American Asp40-allele carriers had a greater cortisol response than Asn40 homozygotes, however, the effect was not observed in participants of Asian descent. In a 12-week open-label trial of 63 Koreans with alcohol dependence, 32 individuals who were medication adherent were genotyped [41]. Half of the sample was Asp40-allele carriers, who were free of relapse for a significantly longer period of time than the Asn40-allele homozygotes (hazard ratio=13.6, p=0.01). Although not statistically significant, Asn40 homozygotes had a 10.6 times greater relapse risk than Asp40-allele carriers. This study was limited by the absence of a control group and the bias introduced by including only medication-adherent patients. A subsequent laboratory study examined whether such findings would obtain in Asian Americans (N = 35; [42]). Using a double-blind, counterbalanced within-subjects design, participants underwent intravenous alcohol administration after receiving treatment with naltrexone or placebo for four days. The procedure was then repeated with participants being switched to the other medication group. Asp40-allele carriers reported less craving for alcohol when treated with naltrexone than placebo during the alcohol administration period. A larger study of the pharmacogenetic effect of subjective response to and self-administration of alcohol in Asian Americans used a similar double-blind design [43], with 77 East Asian participants undergoing two counterbalanced sessions in which they were administered naltrexone or placebo for five days followed by an experimental session in which they received an intravenous priming dose of alcohol followed by a period of alcohol self-administration. Contrary to this group’s prior findings, no pharmacogenetic interaction with medication was observed for either the alcohol-induced subjective response or alcohol self-administration in the laboratory session. In sum, differential effects of OPRM1 based on population have been observed [40, 44], however, those findings are inconsistent, making interpretation difficult.

Laboratory studies, in addition to RCTs, have also been used to evaluate moderation by the Asn40Asp SNP of naltrexone-induced reductions in alcohol consumption. Ray and Hutchison [45] conducted a double-blind, within-subjects alcohol challenge study in which 40 participants genotyped for the SNP were administered placebo or naltrexone and assessed at four ascending breath alcohol concentrations (BrACs): 0.00, 0.02, 0.04, and 0.06 mg/dL. They found that Asp40-allele carriers reported greater blunting of self-reported alcohol-induced high following treatment with naltrexone than placebo, particularly at higher BrACs. A subsequent analysis of the same laboratory study [46] included SNPs in OPRK1 and OPRD1, which encode the kappa-opioid and delta-opioid receptors, respectively, receptors that are also blocked by naltrexone. There is mixed evidenced that variation in these genes is associated with alcohol, opioid use, and cocaine use disorders [47, 48, 49]. The study yielded two gene-by-medication interactions. Following naltrexone treatment, OPRK1 rs997917*T-allele homozygotes reported less alcohol-induced sedation than C-allele carriers and OPRD1 rs4654327*A-allele carriers reported lower levels of alcohol stimulation and craving than G-allele homozygotes. These findings contrast with the earlier null findings for these variation in Gelernter et al.’s [32] analysis of data from the VA Cooperative Study of naltrexone. Setiawan et al. [50] studied 40 “social drinkers,” defined as drinking at least five drinks per week and reporting no alcohol-related problems. Study participants completed an alcohol self-administration task after six days of naltrexone treatment. They found that alcohol-induced euphoria was significantly blunted among Asp40-allele carriers treated with naltrexone, findings similar to those of Ray and Hutchison [51]. In contrast, a laboratory study of 93 participants who completed a cue-reactivity paradigm [52] yielded a gene-by-medication effect that was opposite in direction to these findings. In this study, following 10 days of naltrexone treatment, Asp40-allele carriers reported a heightened urge for alcohol [effect size (η²_p)=0.73], with no effect observed in Asn40-allele homozygotes [52]. The study also examined the effects of a variable number of tandem repeats polymorphism in DRD4, which encodes the dopamine receptor. DRD4 has been implicated as contributing to substance use, attention deficit hyperactivity disorder, and various personality traits [53, 54, 55, 56]. However, it showed no moderating effect (η²_p<0.06) in this study. It is important to note that these laboratory studies were generally conducted in small samples and did not correct for multiple comparisons. Further, the study participants differed among the studies, making it difficult to draw conclusions on the moderating effect of the Asn40Asp SNP on the response to naltrexone.

One approach to addressing the small sample size of pharmacogenetic studies, a limitation that contributes to the risk of a Type 2 error, is to aggregate the effects from multiple studies using meta-analysis. Chamorro and colleagues [57] meta-analyzed six naltrexone treatment studies (N = 764) that evaluated the moderating effect of the Asn40Asp SNP on the likelihood of relapse or abstinence in patients with alcohol dependence. They found that naltrexone treatment in Asp40-allele carriers was associated with lower rates of relapse to heavy drinking than Asn40 homozygotes (OR: 2.02, CI: 1.26–3.22), though there was no evidence of genetic moderation of naltrexone’s effects on the rate of abstinence. A subsequent meta-analysis included eight studies (N = 1,365) in which drinking level was assessed [58]. This analysis showed no significant difference between Asp40-allele carriers and Asn40 homozygotes on the likelihood of a return to heavy drinking during naltrexone treatment.

In summary, evidence of a pharmacogenetic effect of variation in OPRM1 on the response to naltrexone in clinical trials and laboratory studies have yielded mixed results, though the most experimentally rigorous studies have failed to show consistent evidence of an effect. Though beyond the scope of this review, pharmacogenetic studies of naltrexone using functional magnetic resonance imaging have also shown inconsistent findings regarding moderation by the Asn40Asp SNP [37, 59, 60, 61]. This lack of consistent findings is likely due to multiple factors, which are discussed in the Expert Opinion section.

2.1.2. Disulfiram

Disulfiram, the first FDA-approved medication for treating AUD, exerts its therapeutic effects by blocking the metabolism of acetaldehyde, a toxic intermediary metabolite of alcohol. A rapid rise in acetaldehyde occurs when a disulfiram-treated individual consumes alcohol, resulting in an aversive reaction (e.g., flushing, nausea, tachycardia, sweating), the prospect of which is thought to discourage drinking. Although, as described above, DBH rs1611115*T-allele carriers reported fewer drinks per drinking day than rs1611115*C-allele homozygotes [33], there was no moderating effect of OPRM1 on outcomes following disulfiram treatment. A study in 109 Japanese men diagnosed with alcohol dependence who were treated with either disulfiram or placebo for 26 weeks examined the moderating effect of the null ALDH2 allele, which is present in up to half of individuals of East Asian ancestry [62, 63]. Of the 15 participants with this genotype, those treated with disulfiram were more likely to be abstinent at the end of treatment than those who received placebo. To date, there have been no reported efforts to replicate the findings of either of these studies.

2.1.3. Acamprosate

Although the mechanism of action of acamprosate in treating AUD remains to be fully elucidated, and its effects on reducing the risk of relapse to drinking in abstinent patients are modest, the medication is widely available for treating AUD [64]. It was approved in the United States in 2004 based on findings from three European trials that showed it to be efficacious in preventing abstinent patients from relapsing to drinking. It was hypothesized to act via its glutamatergic effects [10]. Ooteman et al. [65] compared the effects of 21 days of treatment with acamprosate or naltrexone on cue-induced craving and examined the moderating effect of SNPs in OPRM1, DRD1, DRD2, GRIN2B, GABRA6, GABRB2, and GABRG2. Although they found that variants in DRD2, GABRA6, and GABRB2 moderated the response to acamprosate, the authors used an alpha level of 0.10 and failed to correct for multiple comparisons. Thus, the results should be seen as preliminary and replication in larger samples is necessary. Karpyak and colleagues [66] examined 548 candidate SNPs in a discovery sample of 225 European-American alcohol-dependent patients evaluated after three months of treatment with acamprosate. They identified two SNPs in GRIN2B (rs2300272 and rs2058878), which encodes a subunit of the NMDA receptor, that has been linked to neurodevelopmental disorders such as intellectual disabilities. They found two nominally significant moderators of abstinence outcomes. Specifically, the minor (G) allele of rs2300272 was associated with a shorter abstinence duration and the minor A allele of rs2058878 was associated with a longer period of abstinence. However, when corrected for multiple comparisons only rs2058878 was significantly associated with the length of abstinence during acamprosate treatment. Although this study did not have a control group, a replication analysis in 110 males from the PREDICT study showed a non-significant trend for an association of abstinence length with rs2058878 (hazard ration=0.72, p=0.07). A prior pharmacogenetic analysis from the three-month PREDICT study implicated rs13273672, a SNP in GATA4, as a moderator of the effects of acamprosate on relapse risk, with G-allele homozygotes relapsing to heavy drinking sooner than the other genotype groups [67]. To date, there have been no published efforts to replicate this finding, though a study of GATA4 variants with alcohol dependence risk showed an association at the gene level, rather than specifically with rs13273672 [68].

2.2. Medications That Are Not Approved in the United States for Treating AUD

2.2.1. Nalmefene

In 2013, nalmefene, a mu- and delta-opioid receptor antagonist, and a kappa-opioid receptor partial agonist, was approved in the European Union for treating AUD, though the literature is not wholly consistent regarding its efficacy [69, 70]. An early study in Finland randomly assigned 403 alcohol-dependent individuals to receive 10–40 mg of nalmefene or placebo on an as-needed basis (i.e., participants were instructed to take the medication when they believed that drinking was imminent) for 28 weeks [71]. The nalmefene-treated group reported fewer heavy drinking days and showed greater improvement on alcohol-related biomarkers than the placebo group. A secondary analysis of data from 272 participants in this study examined the moderating effect of two SNPs in OPRM1, two SNPs in OPRD1 (which encodes the delta-opioid receptor), and one SNP in OPRK1 (which encodes the kappa-opioid receptor) on the response to nalmefene. The study found no main or moderating effects of these genotypes on drinking outcomes [72].

2.2.2. Topiramate

Pharmacogenetic effects of topiramate, an anticonvulsant that is efficacious in the treatment of AUD [73], have also been investigated. Topiramate has multiple pharmacologic effects, including the antagonism of glutamate activity at AMPA and kainate receptors [74, 75] and has been shown to be effective at reducing alcohol consumption, increasing abstinence, and lessening craving [76, 77]. Topiramate’s effects on glutamate receptors are most potent and selective for those containing the GluK1 and GluK2 subunits (encoded by GRIK1 and GRIK2, respectively)[78, 79], which have been linked to schizophrenia, Huntington’s disease, and obsessive-comuplsive disorder, although the findings are inconsistent [80, 81, 82]. To identify a suitable candidate genetic biomarker, Kranzler et al. [83] used an association study in alcohol-dependent cases and controls to examine variation in GRIK1 as a potential moderator of the response to topiramate in treating AUD. They found that, of seven SNPs examined, three were nominally associated with alcohol dependence. Using empirical p-value estimation, only one, rs2832407 (C allele), was significantly associated with the disorder.

Ray and colleagues [84] examined whether the three GRIK1 SNPs identified previously as nominally significant [83] moderated the severity of adverse events in 51 heavy drinkers treated with topiramate or placebo for five weeks. They found that rs2832407*C-allele homozygotes reported a lower severity of adverse events than A-allele carriers. Subsequently, Kranzler and colleagues [85] tested the hypothesis that rs2832407 moderates the response to topiramate in a 12-week, placebo-controlled RCT in heavy drinkers. Analysis of the 122 European-American participants revealed that rs2832407*C-allele homozygotes treated with topiramate reported a significantly greater and more rapid reduction in heavy drinking days and a greater and more rapid increase in abstinent days than both placebo-treated patients and topiramate-treated patients who were heterozygotes or rs2832407*A-allele homozygotes. These investigators did not replicate the previously observed finding that the SNP moderated the adverse effects of topiramate. In a secondary analysis of these data that combined the number needed to treat (NNT) to prevent any heavy drinking in the last month of treatment with the number needed to harm (NNH), i.e., that resulted in a moderate or severe adverse event at any time during the 12-week trial. Among rs2832407*C-allele homozygotes, the NNT for topiramate when adjusted for the NNH was <3, while for A-allele carriers it was >300 [86]. If replicated, these findings would argue strongly for the use of topiramate only in rs2832407*C-allele homozygotes. A post-treatment follow-up of these subjects three and six months after the study medication was discontinued showed that heavy drinking remained low among C-allele homozygotes treated with topiramate [87].

These results, while compelling, require independent replication. Ongoing clinical trials in both the United States (https://clinicaltrials.gov/ct2/show/NCT02371889) and Australia [88] are prospectively randomizing individuals with AUD based on rs2832407 genotype in an effort to replicate the SNP’s moderating effect.

2.2.3. Sertraline

Selective serotonin reuptake inhibitors (SSRIs), which increase serotonin levels in the synapse, are FDA approved for treating anxiety disorders and major depression. In preclinical models, serotonin levels are inversely related to alcohol consumption and in long-term alcohol users there is evidence of serotonin system dysregulation [89, 90]. SSRIs, including sertraline, which facilitates 5-HT transmission while inhibiting dopamine, thus potentially reducing the reward associated with alcohol administration [91], have been tested as treatments for AUD to variable effect [92, 93, 94]. In heterogeneous alcohol dependent populations, these medications have yielded mixed evidence of efficacy, leading to efforts to subtype patient samples to identify both phenotypic and genetic moderators of treatment response [95].

In a study of sertraline for treating AUD [89], 134 participants who were characterized as having either early-onset (i.e., by age 25) or late-onset (i.e., after age 25) alcohol dependence and were genotyped for the tri-allelic serotonin-transporter-linked polymorphic region (5-HTTLPR) polymorphism, which consists of L’ and S’ alleles. This insertion-deletion polymorphism in SLC6A4 encodes the serotonin transporter, which plays a key role in serotonin signaling in the central nervous system. 5-HTTLPR has been linked to alcohol craving [96] and a host of psychiatric phenotypes, including mood, anxiety, suicidality, and the response to trauma [97, 98, 99, 100] Daily alcohol consumption was assessed during the 12-week pharmacotherapy trial and participants were randomly assigned to receive treatment with placebo or up to 200 mg/day of sertraline. The study showed that the effect of sertraline was moderated by age of onset and genotype group, such that among L’ homozygotes, those with late-onset alcohol dependence reported fewer drinking and heavy drinking days when treated with sertraline than placebo. In contrast, L’ homozygotes with early-onset AUD who were treated with placebo reported fewer drinking days and heavy drinking days than those receiving sertraline. There were no effects in S’-allele carriers. As noted by the authors, there was high rate of attrition in this study, particularly amongst early-onset L’ homozygotes, and the results require replication in larger samples.

Kenna et al. [91] examined the pharmacogenetics of two serotonergic medications in a randomized, crossover laboratory study of 77 non-treatment-seeking individuals with alcohol dependence. Participants were randomly assigned to receive three weeks of treatment with sertraline 200 mg/day or ondansetron 0.5 mg/day, placebo for three weeks, and followed by the alternative medication (sertraline or ondansetron) for three weeks. Each three-week treatment period was followed by a laboratory assessment. They examined the moderating effect of the bi-allelic 5-HTTLPR polymorphism (which yields SS, SL, and LL genotypes) and found that women with an S allele drank fewer drinks per drinking day than L-allele homozygotes when treated for three weeks with sertraline. The results for men were not significant. There was also a three-way interaction of 5-HTTLPR and DRD4 genotypes and medication, such that S-allele carriers with the DRD4<7-repeat allele treated with sertraline drank less during the laboratory session and in the naturalistic period for the duration of the trial. These complex preliminary findings in both a clinical trial and a laboratory study require replication before they can be used to inform the pharmacogenetics of AUD treatment.

2.2.4. Ondansetron

Ondansetron, a 5-HT₃ receptor antagonist that is approved as an antiemetic [101, 102], has also been evaluated for the treatment of AUD. The medication has been shown to reduce craving, lengthen periods of abstinence, and reduce heavy drinking in participants with AUD in whom it may reduce the rewarding effects of alcohol [64, 101]. Johnson et al. [103] randomly assigned 283 alcohol-dependent individuals to receive 11 weeks of treatment with ondansetron 4 μg/kg twice daily or placebo, together with cognitive behavioral therapy. The randomization was stratified using the bi-allelic 5-HTTLPR polymorphism and participants were also genotyped for a SNP (rs1042173) in the 3’ untranslated portion of the serotonin transporter gene, SLC6A4. Results indicated that among L-allele homozygotes, those treated with ondansetron reported fewer heavy drinking days, fewer drinks/drinking day, and higher abstinence rates than those treated with placebo. There was no effect of ondansetron among S-allele carriers. Finally, there was a gene-gene interaction, such that 5-HTTLPR*L-allele homozygotes who were also rs1042173*T-allele homozygotes reported consuming the fewest drinks per drinking day and having the highest percent days abstinent of the four genotype-by-medication groups. A secondary analysis of these data was conducted that included 19 SNPs in HTR3A and HTR3B, the genes encoding the 5-HT3 receptor, a key binding site for ondansetron [104]. It identified an additional three genotypes that moderated the response to ondansetron. Using these SNPs and the two previously identified polymorphisms in SLC6A4, it was possible to categorize 34% of European-Americans as having at least one of the five ondansetron-responsive genotypes. Further, individuals possessing any combination of three identified HTR3A or HTR3B genotypes had a five-fold likelihood of having no heavy drinking days when treated with ondansetron compared to placebo. The authors referred to this subgroup as “super responders” [104].

Kenna et al. [91], using the crossover study design described above, examined the pharmacogenetics of the response to ondansetron [105]. They found that ondansetron modestly reduced naturalistic drinking in 5-HTTLPR L-allele homozygotes, but the effect was not seen in the laboratory self-administration part of the trial.

Overall, the initial findings of beneficial effects of ondansetron on drinking in the naturalistic setting are promising, as the medication exerted these effects at a very low dosage, at which it is well tolerated. An ongoing trial (https://clinicaltrials.gov/ct2/show/record/NCT02354703) seeks to replicate and expand upon the findings reported by Johnson et al. [104].

2.2.5. Baclofen

Baclofen, a GABA_B receptor agonist, was recently approved for AUD treatment in France, though the findings from RCTs are not consistent in supporting its efficacy for that indication [16]. Some of the heterogeneity in the response to baclofen may be due to the presence of psychiatric comorbidity in some study populations [21]. In view of the mixed findings, further research on its use is required, leading to a recent consensus statement recommending that it be considered a second-line pharmacotherapy for AUD [106]. In the only published pharmacogenetic study of baclofen, Morley and colleagues [107] analyzed data from a subset of alcohol-dependent participants who were randomly assigned to treatment with baclofen (30 or 75 mg/day) or placebo. In the secondary analysis, they examined the moderating effect of rs29220 in GABBR1, a subunit of the GABA_B receptor gene in 72 subjects. GABA_B impacts reward signaling and has been implicated in various psychiatric and neurological disorders including schizophrenia, substance use disorders, and epilepsy [108, 109, 110, 111]. They found that rs29220*C-allele homozygotes treated with baclofen reported a longer time to relapse and greater proportion of abstinent days than those treated with placebo or baclofen-treated participants with one or two rs29220*G alleles. They also explored the moderating effects of rs29230 of GABBR1 and rs7865648 of GABBR2 (which encodes a second GABA_B receptor subunit), but found no significant effects of these SNPs. They concluded that variation in the rs29220 allele frequency by population may help to explain the lack of efficacy of baclofen in some studies. Further study of this potentially useful medication, along with genetic predictors of treatment response, appears warranted.

3. Expert Opinion

In sum, the literature on the pharmacogenetics of AUD treatment is not yet adequate to inform clinical practice. There remains a paucity of pharmacogenetic trials in the field, which is not surprising in that overall the literature supporting medication efficacy in AUD is limited, as is our understanding of the mechanism of action of these medications [24, 112]. Although initial studies have yielded evidence of a number of promising genetic moderators of pharmacotherapeutic response, many of these studies have been conducted in small samples, potentially yielding false positive or false negative findings. In addition, some studies failed to correct for multiple comparisons or to conduct intent-to-treat analyses and were inconsistent in their selection of participants’ level of alcohol consumption, severity of AUD, and demographic features. These methodological shortcomings limit the interpretation of these studies’ findings and make comparison across studies difficult. Prospective studies, the optimal approach to validating a pharmacogenetic hypothesis, when they have been attempted, have yielded null or clinically insignificant effects. Until there exists a more extensive set of studies in which individuals with AUD are prospectively randomized to active and placebo conditions using an empirically-based pharmacogenetic moderator, the selection of medications to treat the disorder will remain largely a matter of trial and error.

Because AUD is a heterogeneous disorder, variability in sample characteristics complicate efforts to identify a moderating effect of genetic variation. Treatment-seeking status, for example, has a potentially important effect on outcomes in alcohol treatment research [113, 114], as heavy drinkers may be drawn from a different population than individuals with an established AUD diagnosis. Oliver and McClernon [115] argued that attention should also be paid to the timing of the intervention and the temporal course of the disease being treated rather than considering the treatment or its genetic moderation as static features. Psychosocial interventions, including medication management [116], are commonly used in conjunction with medications. Thus, they are an important factor in determining treatment outcomes, as the combination of medication and psychosocial therapy potentially results in better outcomes than either alone [117].

The studies discussed in this review used a wide range of outcome measures, including good clinical outcome [31], subjective response [43], and self-reported alcohol use [36, 85], which limits comparisons across trials and the interpretation of divergent findings. This highlights the need for cost-effective, reliable, and valid biomarkers for alcohol use [118, 119], which could standardize outcome assessments and increase their validity. The size of a standard alcoholic beverage differs across countries, making cross-national comparisons between studies more difficult. The reliance on self-reported alcohol use as an outcome measure, in addition to likely providing an underestimate of drinking, fails to account for improvements in psychosocial functioning and quality of life, important goals of treatment, which have been shown to improve with even small reductions in alcohol consumption [120]. Endophenotypes, such as the subjective response to alcohol [121] or sub-phenotypes, such as the type of reinforcement drinking that characterizes participants [122], could be tested as treatment targets or incorporated in models of treatment outcome, given their greater proximity to underlying genetic variation and the neurobiological changes that result from alcohol use. A heretofore neglected consideration in the design of pharmacogenetic RCTs is the extent to which the placebo effect is heritable [123, 124]. Publication bias, including the “file drawer” effect in which negative or null studies are not published [10], could also contribute to difficulties in interpreting the pharmacogenetic literature.

Although several of the SNPs studied as genetic moderators of AUD treatment have strong biological plausibility, we do not yet fully understand the function of many genes or the impact of variation in them, which limits opportunities for discovery and the interpretation of pharmacogenetic findings [125]. It is possible that, due to the polygenic nature of AUD, individual polymorphisms, such as the Asn40Asp SNP in OPRM1, exert small effects on the development and maintenance of AUD without having a clinically meaningful moderating effect on the response to pharmacotherapy [125].

Numerous genome-wide association studies have identified variants in the aldehyde and alcohol dehydrogenase enzyme genes that can substantially affect alcohol metabolism [126, 127, 128]. Whereas most of the medications reviewed here (e.g., naltrexone, ondansetron, and sertraline) are largely metabolized in the liver, the effects of genetic variation on the pharmacokinetics of these medications is also of potential importance in advancing the precision treatment of AUD [129].

Several practical limitations exist in conducting research on the pharmacogenetics of AUD treatment. Other specialty areas, such as oncology, are much further along in the development of this and other aspects of precision medicine [130]. Contributing to this disparity is the disproportionately greater funding provided to the National Cancer Institute, for example, than the National Institute on Alcohol Abuse and Alcoholism, despite the high prevalence and social impact of AUD and its causal relation to a variety of cancers [130]. As a result, there are well understood pharmacogenetic approaches to tumor treatment that are now widely used clinically. For example, tumors characterized by mutations in EGFR that cause overexpression of the EGFR protein are linked to a poorer prognosis in several types of cancer. Treating individuals whose tumors carry the mutation with protein kinase-inhibiting drugs has been shown to increase response rates from 10% to 75% [130, 131]. Unfortunately, addiction medicine has no such dramatic examples of pharmacogenetic effects.

Because AUD is associated with a variety of physical and mental health disorders, progress in understanding the etiology and advancing the treatment of AUD can yield benefits in other areas of medicine. Once valid moderators of pharmacotherapy for AUD are identified and validated, disseminating and implementing them in practice will require a concerted effort to educate providers, many of whom still do not routinely prescribe medications with demonstrated efficacy in treating the disorder [16, 132]. A contributor to the low rate of medication use to treat AUD [13, 14] may be resistance from patients. Thus, patients will need to be educated in the utility and clinical relevance of both pharmacotherapy and genetics, before these can assume a central role in the treatment of AUD. Further, the implementation of pharmacogenetics in clinical settings will require staff training, logistical support (e.g., updating electronic health record systems and laboratories), enhanced cybersecurity, and changes in policies regarding insurance coverage.

The revolution in genetics that was ushered in by the sequencing of the human genome will benefit precision medicine in all areas of medicine, including the pharmacogenetics of AUD treatment. Recent strides have been made in recruiting large samples for genetics studies, which have yielded databases that are useful for the discovery of variants contributing to the risk of disorders and have begun to be interrogated to identify moderators of treatment. Genomic and phenotypic data from the UK BioBank [133, 134], which recruited 500,000 participants from the general population in the United Kingdom, are now publicly available. The Million Veteran Program [133], an effort of the U.S. Department of Veterans Affairs, has now recruited more than 700,000 veterans for whom genotypic and phenotypic information from questionnaires and the electronic health record [135], represents a valuable source of data for discovery. These efforts have begun to yield novel genetic loci implicated in alcohol consumption and alcohol-related problems [136] and more can be expected in the coming years.

Pharmacogenetic studies of AUD will benefit from larger, more diverse, and better characterized samples that can be followed longitudinally. Elucidation of the pathophysiology of AUD and identifying valid endophenotypes will enhance the discovery of genetic moderators of AUD treatment and allow the targeting of genetic and biobehavioral mechanisms. Ultimately, though, stakeholders (including basic and clinical researchers, clinicians, and clinical administrators must align their goals and resources [118, 130] to realize the fundamental goals of precision medicine as applied to AUD treatment.