Five Priority Areas for Improving Medications Development for Alcohol Use Disorder and Promoting Their Routine Use in Clinical Practice

ALCOHOL USE DISORDER (AUD) is a complex and devastating disorder, resulting in a myriad of medical, psychological, social, and economic problems. Harmful alcohol drinking costs the U.S. society more than $249 billion annually and is the 5th leading risk factor for premature death and disability. Each year, more than 15 million Americans are diagnosed with AUD and 88,000 die from alcohol-related causes. Globally, those numbers are staggering, with 3.3 million deaths attributed to alcohol each year (https://pubs.niaaa.nih.gov/publications/AlcoholFacts&Stats/Alcohol Facts&Stats.htm). During the past 2 decades, advances have been made to develop medications to treat AUD, as evidenced by the U.S. Food and Drug Administration’s (FDA’s) approval of disulfiram, oral and long-acting injectable naltrexone, and acamprosate for alcohol dependence. In addition, nalmefene was approved in Europe and baclofen was approved in France for the treatment of alcohol dependence. Other medications are not approved but have shown efficacy in some but not necessarily in all clinical trials. These medications include topiramate, varenicline, ondansetron, gabapentin, aripiprazole, and prazosin/doxazosin (Witkiewitz et al., 2019a).

Although these medications have exhibited efficacy in some people and trials, none of the medications work for everyone consistently in all trials. This is due to the heterogeneity of AUD (e.g., variations in genetics and environment), which leads to a differential response to medication. As a result, clinical trials of these medications show only small effect sizes (Litten et al., 2016a). The small effect sizes shown by both approved medications and those recently investigated in clinical trials contribute to the fact that clinicians may be skeptical and reluctant to prescribe medications for AUD. It is estimated that less than 4% of patients with AUD are prescribed FDA-approved medications (Litten et al., 2016b).

The 2 biggest challenges over the next decade will be to (i) discover and develop medications that would have a larger effect size in alcohol pharmacotherapy clinical trials and (ii) facilitate the use of medications in real-world clinical settings. In addressing these challenges, it is important to understand the drug discovery process. Drug development takes a long time (approximately 18 years for CNS compounds) to move a potential medication from its initial discovery phase to the marketplace, often costing nearly 2 billion dollars (Kaitin and Milne, 2011). The slow pace and high costs are due to the failure of many compounds; only 8% of new CNS compounds entering Phase I studies will reach the market (Kaitin and Milne, 2011). It is essential to make the drug development process more efficient in terms of speed, cost, and predictability (Litten et al., 2016b). Likewise, we need to better understand the obstacles that stand in the way of prescribing AUD medications once they are available. If we are to improve the development and use of AUD medications, progress needs to be made at least in 5 key areas. These include: (i) identifying targets that yield more effective candidate compounds, (ii) advancing personalized medicine, (iii) developing screening models that predict a high probability of clinical success, (iv) engaging pharmaceutical companies in medications development for AUD, and (v) developing strategies to better incorporate alcohol treatment medications in every day clinical practice. In this article, we describe strategies for advancing these 5 key areas (Table 1).

Table 1.

Five Priority Areas for Improving Medications Development for AUD and Promoting Their Routine Use in Clinical Practice

1. Identify More Effective Druggable Targets

• Identify single targets that alter drinking

• Identify downstream changes from single target activation (e.g., Connectivity Map method)

• Mine data from biomolecular and cellular networks

2. Advance Personalized Medicine

• Measure demographics and background information (e.g., family history, AUD severity, comorbidity, drinking patterns)

• Assess self-reports and neuropsychological measures

• Assess comorbidity with substance use, mental health, and/or medical disorders

• Measure objective biomarkers

o Integration of-omics

o Imaging

o Electrophysiological measures

o Cell and tissue model systems

• Define and measure addiction domains

• Explore single-patient approach (e.g., n-of-1 trial)

3. Develop Screening Models that Predict High Probability of Clinical Efficacy

• Link screening models to clinical trials, including bidirectional approaches between animal and human laboratory models and clinical trials

• Create high-throughput screening models

• Explore cellular and tissue model systems

• Investigate small vertebrate and invertebrate model organisms

4. Engage Pharmaceutical/Biotechnical Companies in Medications Development for AUD

• Encourage collaboration of government/academia with pharmaceutical/biotechnical companies

• Share resources of pharmaceutical/biotechnical companies with government/academia (e.g., library of novel compounds, finance expensive Phase 3 trials, develop and implement extensive marketing of their FDA-approved compounds)

• Share resources of government/academia with pharmaceutical/biotechnical companies (knowledge of druggable targets for AUD, development of alcohol animal and human laboratory paradigms, knowledge of conducting alcohol treatment clinical trials)

5. Facilitate the Use of Alcohol Treatment Medications in Clinical Practice

• Increase efforts by NIAAA, other federal agencies, and professional organizations to reduce barriers to medication use

• Implement medical education in addiction medicine

• Reduce stigma

• Make screening, intervention, and treatment of AUD a routine practice in primary care and specialty medical settings

IDENTIFY MORE EFFECTIVE CANDIDATE COMPOUNDS

During the past 3 decades, alcohol clinical trials have been conducted on more than 30 different medications, exploring a wide range of potential molecular targets. Most of these studies showed either no significant difference from placebo or only small effect sizes (Jones et al., 2014; Litten et al., 2016a; Zindel and Kranzler, 2014). In recent years, even more targets have been identified as we gain a better understanding of the mechanisms underlying alcohol-seeking and drinking behavior (Akbar et al., 2018; Koob and Mason, 2016; Litten et al, 2016a; Yardley and Ray, 2017). These targets typically are identified using animal models, an approach that has remained generally the same over the past 30 years. This approach has been effective, yielding a menu of medications, but all remain in the small effect-size range.

The alcohol field is not the only one to face this situation. Other fields with complex heterogeneous disorders—such as cancer, cardiovascular disorders, other drug addiction and psychiatric disorders like depression, posttraumatic stress disorder, and other anxiety disorders—have experienced similar outcomes. Increasing the effect sizes of medications to treat these complex disorders requires new strategies that not only focus on identifying a single target, but also consider what happens downstream from that target such as how the target interacts with other targets and pathways (Ferguson et al., 2018a). Examples of such strategies are described below.

Connectivity Map (cMap) methodology explores the actions of medications across different pathways and systems. Scientists use this technology to create a gene expression profile, which is fed into a database to map the molecular signatures associated with a particular disorder, such as AUD. Those profiles then are compared with multiple compounds to identify ones that might prevent or alleviate the disease (Lamb et al., 2006). The best candidates are ones that produce changes in gene expression opposite to those associated with the disease state (Kunkel et al., 2011; Lamb et al., 2006; Subramanian et al., 2017). This methodology is being used to identify new compounds in a variety of diseases, including cancer, skeletal muscle atrophy, Parkinson’s disease, schizophrenia, and Alzheimer’s disease (Ferguson et al., 2018a; Kunkel et al., 2011). The National Institutes of Health (NIH) has established a Library of Integrated Network-Based Cellular Signatures (LINCS), a program that collects molecular signatures based on gene expression caused by various diseases, with a collection of more than 16,000 compounds (https://commonfund.nih.gov/LINCS/). Using this method, we are making progress toward finding new compounds for treating AUD. Ferguson et al. (2018b) discovered several compounds that scored high for potential repurposing, including pergolide, terreic acid, genipin, alvespimycin, and BRD-K14355517, a neuropeptide S antagonist. The large number of compounds in the LINCS program has given us many new compounds to explore and requires a high-throughput screening model to further refine the best possible candidates (see Screening section below).

Another approach for discovering druggable targets for AUD is the development, integration, and data mining of biomolecular and cellular networks (Hopkins, 2008; Masoudi-Nejad et al., 2013; Yildirim et al., 2007). Instead of focusing on one target, this approach determines how targets are related to one another and how they correlate to changes underlying AUD. Several networks exist, including gene–gene, gene–protein, protein–protein, and metabolic and regulatory networks (Gebicke-Haerter, 2016; Robinson and Nielsen, 2016). These networks of genes, proteins, and metabolites are responsible for how neurons function and how they are connected to form the neurocircuits that are integral to the different domains of AUD. Although highly complex, many researchers believe that knowing how these networks work is essential, not only for identifying which targets drive AUD but in developing effective treatments. Researchers are using this approach to examine a range of complex disorders, including cancer, mood disorders, schizophrenia, endocrine disorders, and Huntington’s disease (Collier et al., 2016; Morrow et al., 2010; PIrhaji et al., 2016; Yildirim et al., 2007). The National Institute on Alcohol Abuse and Alcoholism (NIAAA) is taking steps to facilitate this line of research in the alcohol field, including issuing a guide, “Development, Integration, and Data Mining of Biomolecular and Cellular Networks for Discovering Druggable Targets for Alcohol Use Disorder and Alcohol-Induced Organ Damage” (https://grants.nih.gov/grants/guide/notice-files/NOT-AA-17-007.html).

Human cell- and tissue-based model systems offer alternative platforms for discovering medication targets. Induced pluripotent stem cells (iPSC) enable researchers to develop tissue-specific cell-based disease models (e.g., neurons, glial cells, etc.) for disorders like AUD (Horvath et al., 2016; Inoue et al., 2014; Ko and Gelb, 2014). This approach, sometimes called the “organ-on-a chip” method, uses stem and progenitor cells to differentiate into multiple cell types that represent the cellular architecture within the organ, thus modeling a system for drug screening (https://ncats.nih.gov/tissuechip/projects) This microphysiological system enables us to explore the activities, mechanics, and physiological responses of a functioning organ through microfluidic culture devices that recapitulate the structure and function of living the human organ. This microfluidic assay system combines tissue and matrix substrates (Horvath et al., 2016). Advances in 3-dimensional (3D) cell culture technology also offer other new ways of mimicking the complex in vivo biology and serve as new tools for drug discovery. In addition, 3D technology is being combined with “cell painting”—a new, high-content, image-based cellular profiling technique—to examine similarities and differences among cells from various disease states (Bray et al., 2016; Pennisi, 2016). With this method, scientists can assign 6 fluorescent dyes to reveal up to 1,500 morphological features, perturbations of which can be used for drug screening. Tissue chips and organoids derived from the cells of people with AUD could lead to identification of potential treatment targets, particularly if the targets can reverse or prevent changes associated with AUD.

Gaining a better understanding of how the brain works and how AUD changes brain function will likely produce the fundamental knowledge for the drug development of AUD. In addition, new approaches and methodologies will enhance our ability to discover medications and hopefully to identify more effective druggable targets. Because of the complexities of AUD, multiple targets contribute to the disease. Exploring combinations of targets using network analysis and machine learning may help to produce synergistic drug combinations that are both effective and safe.

ADVANCE PERSONALIZED MEDICINE

Advancing personalized medicine (or precision medicine) may also improve treatment outcome with medications. Because of the heterogeneous nature of AUD, it is not realistic to develop one medication that is successful for everyone who suffers from this disorder. The goal of personalized medicine is to identify and target specific phenotypes that are most likely to respond favorably to a given medicine. Over the past 20 years, alcohol treatment trials designed to match a person with the medication most likely to be effective for him or her has been challenging. Individual characteristics—such as demographics, family history, AUD severity, and variations in assessment (through self-reports and neuropsychological instruments)—have complicated our ability to identify subjects who will respond best to a particular experimental medication. One reason may be the lack of objective biomarkers. There has been some progress in biomarkers, especially pharmacogenetics, although this research is still in its early stages (Garbutt et al., 2014; Jones et al, 2015; Sun et al., 2016). Some promising studies include one by Kranzler and colleagues (2014), who found that AUD patients with a CC genotype of the rs2832407 GRIK 1 gene encoding the glutamate kainite GluK1 receptor responded better to topiramate than those with the other genotypes (AC and AA genotype). These results are promising but obviously in need for a replication which is in fact one of the aims of a prospective clinical trial recently completed and whose data analysis is underway (NCT 02371889). A pharmacogenetic approach has also been proposed in studies testing the 5-HT3 antagonist ondansetron in AUD individuals. Specifically, Johnson et al. (2011, 2013) reported a pharmacogenetic interaction between ondansetron and the SLC6A4 gene (which encodes the serotonin transporter [5-HTT]) at the L versus S alleles and the rs1042173 (T vs. G), and the HTR3A (rs17614942 and rs1150226) and HTR3B (rs17614942) genes. Notably, a human laboratory study also supports the interaction of ondansetron with the SLC6A4 gene polymorphism on alcohol-related outcomes (Kenna et al., 2014). The single-nucleotide polymorphism (A/G rs179997) in the u-opioid receptor gene (OPRM1) has been proposed to influence naltrexone responses in AUD patients (Anton et al., 2008; Garbutt et al., 2014; Oslin et al., 2003), but the results have been mixed, for example, a recent prospective study showed no effect in treatment response to naltrexone in patients with the OPRM1 A/G polymorphism (Oslin et al., 2015). It is not unusual that prospective studies do not always replicate retrospective results because, in contrast to prospective studies, unbalanced groups usually occur in retrospective results. So far, none of these pharmacogenetic findings are ready to be implemented in real-world practice (Hartwell and Kranzler, 2019). Beyond pharmacogenetic approaches, clinically driven subgroups may help in guiding more personalized approaches, such as identifying an AUD medication based on a specific patient’s comorbidity, that is, a substance use disorder, mental health and/or medical disorder. For example, clinical trials suggest that varenicline may work best in AUD patients who are also smokers (Falk et al, 2015; Litten et al, 2013.). Similarly, while results with aripiprazole, prazosin/doxazosin, nalmefene, and baclofen have been conflicting, clinical studies suggest that aripiprazole may be effective in AUD patients with high impulsivity and/or low self-control (Anton et al, 2017); prazosin/doxazosin may help patients with AUD and high blood pressure (Haass-Koffler et al., 2017; Wilcox et al, 2018); nalmefene may help patients with very high drinking risk levels (Miyalta et al., 2019); and baclofen may work best in those with higher baseline alcohol consumption, clinically significant liver disease, and those who were CC homozygous carriers of the GABBR1 rs29220 polymorphism (Addolorato et al., 2007; Agabio et al., 2018; Leggio and Lee, 2017; Morley et al., 2018a, 2018b; Pierce et al., 2018; Rombouts et al., 2019). Finally, it has been shown that naltrexone works favorably in those who report drinking alcohol for its rewarding or pleasurable effects rather than as a means of relieving negative experiences (high reward/low relief) (Mann et al., 2018; Witkiewitz et al., 2019c).

Despite these advances, and given the complex nature of AUD, it is unlikely that one single factor will be enough to paint a complete picture of who will respond to a given medication. In addition to patients’ characteristics and self-reported information, a multiple-pronged approach likely will be needed to collect objective biomarkers associated with the disease, treatment response, and phenotype. Methodological and computational analytical approaches are being developed to integrate data from multiple “omics” studies (genomic, transcriptomic, metabolomic, and proteomic profiles) (Redon and Monleon, 2015) as well as information from various imaging techniques, including functional magnetic resonance imaging (fMRI) studies (e.g., cue reactivity fMRI paradigms), positron emission tomography (PET) studies, transcranial magnetic stimulation (TMS)-guided electroencephalogram (EEG), and/or other brain stimulation approaches (Belardinelli et al., 2019; Bach et al., 2019; Diana et al., 2017; Etkin et al., 2019; Heilig and Leggio, 2016; Mann et al., 2014; Ilmoniemi and Kicic, 2010; Salling and Martinez, 2016). Machine- and deep learning-based algorithms could be developed that incorporate all these biomarkers—along with other patient characteristics (e.g., demographics, family history, disease severity, pattern of drinking, genetic and epigenetic markers, medical and/or psychiatric comorbidities)—to predict medication response.

Cell and tissue model systems for identifying effective druggable targets (described above) also have applications in personalized medicine (Horvath et al., 2016; Inoue et al., 2014). In fact, this approach already is being used with other complex diseases. For example, in cancer, cellular models are being used to treat breast cancer (Yu et al., 2018). Yu and colleagues used this technique to create a patient-derived tumor xenograph and tumor organoids to treat a specific type of breast cancer. Using this procedure, they identified a new type of medication, called a DNA methyltransferase inhibitor, which successfully reduced the tumor size. Generating personalized cellular tissue through in vitro human-induced pluripotent stem cells (hiPSC)-based models is now being considered as a treatment for psychiatric disorders (Brennand, 2019). Recently, Ho and colleagues (2019) showed that alcohol and acamprosate alter tetraspanin 5 (TSPAN5), a gene that alters serotonin levels in hiPSCs that were differentiated into astrocytes and cortical neurons.

In addition to the approaches above, it is important to identify the different domains of AUD and their corresponding brain circuits (Litten et al, 2015). To this end, NIAAA initiated the Addictions Neuroclinical Assessment (ANA) program, which defines 3 relevant functional domains of AUD: incentive salience, negative emotionality, and executive function (Kwako et al., 2016, 2017). The 3 domains are assessed using a battery of behavioral and physiological tasks, blood tests for genetic analysis, and self-report measures. Integrating these measurements could help us determine how the 3 domains differ among individuals with and without AUD, how they vary within the AUD population as a whole, and, most importantly, how they can be used as targets/subpopulations for medications development.

Because of the complexities of AUD and the different responses to treatment among patients, it has also been suggested that personalized medicine should be studied through single-patient (N-of-one) trials (Davidson et al., 2014). The goal of the n-of-one trial is to determine the optimal or best intervention for an individual patient using objective data-driven criteria (Lillie et al., 2011). Early results using this method indicate that it might be useful in treating depression (Kronish et al., 2018). The concept of measuring individual behavioral responses is evolving. For example, a “just-in-time” approach measures how individuals differ in their behavioral responses to interventions applied during periods of particular risk for a given individual (Liao et al., 2019).

In summary, many exciting possibilities exist in personalized medicine over the next decade. In addition to simply measuring the difference between the experimental and placebo groups, clinical studies need to focus more on distinguishing between responders and nonresponders. To advance this concept, several goals must be attained. Fully accessible biobanks and data repositories that allow researchers to share their findings will be important in reaching these goals. As much data as possible should be collected from the subjects. This includes collecting samples across the -omics sciences and iPSC as well as imaging studies for objective markers. To this end, NIAAA recently launched a major data-sharing initiative whereby human subjects data collected from most NIAAA-funded grants will be submitted to a data repository called the NIAAA Data Archive (https://grants.nih.gov/grants/guide/notice-files/NOT-AA-19-020.html). With time, the NIAAA Data Archive will eventually contain large amounts of complex data, standardized to facilitate analysis. The next steps will involve developing and applying novel computational analytical approaches, such as machine and deep learning technologies, to look for trends in how people respond to different medications (Belle et al., 2013; Ching et al., 2018; Webb, 2018). Linking clinical trial researchers to data/computational/modeling analysts will be an important step in helping achieve this goal. This is a clear example of how medication development for complex medical disorders, such as AUD, is best accomplished when scientists from many disciplines work together as a team. This multidisciplinary approach also suggests the need to grow, mentor, and train not only new generations of basic neuroscientists, clinical researchers, physician-scientists, and epidemiologists, but also the need for a larger group of new scientists trained in bioinformatics, big data, and computational neuroscience.

DEVELOP SCREENING MODELS THAT PREDICT HIGH PROBABILITY OF CLINICAL EFFICACY

Drug development is expensive and time-consuming, especially for those drugs that target the central nervous system. It takes a long time to develop a medication, from the initial discovery of a potential molecular target to the time the final compound is available in the marketplace. The process is plagued by setbacks. Many promising drugs fail due to a lack of efficacy and/or safety issues, often in the final stages of the drug development, in the costly clinical trial phase, or even after approval (Phase 4 pharmacovigilance)—after a considerable investment of time and resources (Paul et al., 2010). One way to improve the efficiency of the drug development process is to develop better screening models that preempt having to conduct more expensive and time-consuming clinical trials of compounds that ultimately will fail. If we can be more accurate in predicting who is most likely to benefit from a medication, we would have a higher probability of positive outcome for clinical trials. This approach will also help ensure we do not expose individuals to medications under development that may lack efficacy and/or lead to issues related to safety and tolerability. In the alcohol field, where animal models and human laboratory paradigms are used for screening participants, we have had only mixed results in predicting a positive clinical outcome for a given medication (Egli, 2018; Yardley and Ray, 2017). Preclinical studies were successful in predicting the effectiveness of naltrexone and acamprosate, both of which were approved by the FDA for alcohol dependence. But a major “false-positive” disconnect exists with other medications, such as SSRIs, bromocriptine, and memantine, whereby these compounds showed promise in animal studies but lacked efficacy in clinical trials (Egli, 2018; Yardley and Ray, 2017). Nonetheless, using screening models to understand the mechanism of action of the candidate compound should help inform the best design and population selection for clinical studies.

A variety of alcohol animal models and experimental paradigms are currently being used to reflect the different stages of the addiction cycle: binge–intoxication, withdrawal–negative affect, and preoccupation–anticipation (Bell et al., 2012; Egli, 2005; Koob and Mason, 2016). For example, the NIAAA Standardized Animal Model Program is an initiative consisting of a network of sites used to test promising candidate compounds (Litten et al., 2016b). The models include rats bred to seek and consume high levels of alcohol (up to 2 bottles), referred to as alcohol preferring (P) rats and high-alcohol-drinking-1 (HAD1) rats (contract with Indiana University), and mice and rats exposed to EtOH vapor that mimic dependency in humans (contract with Medical University of South Carolina and with Louisiana State University School of Medicine). By standardizing the use of these models at each site, NIAAA hopes to limit the variability (in both methods and in animal species and strains) so often found across sites. To date, this program has evaluated many novel compounds and reference medications (compounds whose efficacy has already been tested in human clinical trials). Because the results from the clinical trials are already known, evaluating the efficacy of the reference compounds in animal models is especially useful to validate the predictive utility of these models.

In addition to animal models, several human laboratory paradigms exist that can be used to screen candidate compounds for efficacy as well as to determine their mechanisms of action (Bujarski and Ray, 2016; Plebani et al., 2012). Similar to the animal model program, NIAAA recently initiated a Human Laboratory Program (HLAB) to standardize human laboratory paradigms to serve as screening models for candidate compounds. The first study used multiple sites to examine varenicline (NCT: 03035708). A second study began in the summer of 2019 with the novel compound ANS-6637 (an aldehyde dehydrogenase-2 inhibitor and dopamine inhibitor) (NCT: 03970109).

These animal and human laboratory screening models should help us make more rational Go/No-Go decisions in advancing candidate compounds through the drug development process. The models hinge on our ability to bridge preclinical animal models and clinical human laboratory models. At the very least, bidirectional validation should be performed both to compare the results of the animal and human laboratory models for a given candidate compound and to compare the results of the screening models with those of clinical trials (Litten et al., 2016b).

An often recurring question in the alcohol field is whether animal models can reliably predict treatment success in humans. Undoubtedly, linking animals to complex human behaviors is challenging. In fact, animal models have rarely predicted success in clinical trials for CNS disorders, such as depression, anxiety, and schizophrenia (Hyman, 2012). The results in the alcohol field are not so clear, with studies showing mixed results (Egli, 2018; Heilig et al, 2016; Ray et al., 2019). It may be that we simply need to explore additional explanations. For example, we know that AUD is a heterogeneous disorder and treatment medications appear to work better in certain subtypes of AUD patients. Perhaps, the best role for animal models is to not only inform the type of paradigm to be used, but also to refine the selection of candidates for a human laboratory study to maximize the chance of matching the right medication with the right individual. For instance, if the experimental medication reduces stress-induced drinking behavior in animals, it seems likely that the medication would work effectively in humans who have a hyperactive stress system. A recent clinical trial on vasopressin 1b antagonist, which has been hypothesized to reduce stress, appears to support this hypothesis; in a post-hoc analysis, results showed a better drinking outcome in patients reporting high stress levels compared with subjects who did not report high stress (Ryan et al., 2017). Notably, animal data also support the potential role of the V1b receptor as a treatment target. Specifically, the V1b receptor antagonist SSR149415 reduces alcohol drinking in Sardinian preferring rats (Zhou et al., 2011) and in alcohol-dependent rats (Edwards et al., 2012). Animal paradigms might give insight into which type of AUD patients should be included in human studies and which human laboratory paradigms are likely to lead to the most informative results.

Finally, in seeking new approaches for identifying compounds, it may be necessary to develop high-throughput screening models to test multiple candidate compounds and drug combinations, especially before testing in animal models. For example, the Connectivity Map approach, described earlier, can be used to identify a number of potential medications. Narrowing those choices using high-throughput screening could save time and resources and lead to a more efficient drug development process. Advances also are being made in developing both in vitro and ex vivo models. For example, iPSCs, described above, hold promise for drug discovery. This technology has been used to identify compounds that show relevance for specific diseases, like amyotrophic lateral sclerosis, Alzheimer’s disease, familial dysautonomia, Rett syndrome, schizophrenia, spinal muscular atrophy, as well as AUD (Inoue et al., 2014; Ko and Gelb, 2014; Prytkova et al., 2018). Organoids grown from iPSC can be used to build 3-dimensional models of tissue, such as different parts of the brain (mini-brains) (Shen, 2016; Simm et al., 2018). Those models can be further refined through cell painting, an imaging assay that uses multiplexed fluorescent dye to profile individual cells (Bray et al., 2016; Dorval et al., 2010; Goodman and Carpenter, 2016; Gustafsdottir et al., 2013; Kokel et al., 2010; Pennisi, 2016; Ranson et al., 2019;).

Small vertebrate and invertebrate model organisms can also be used for drug high-throughput screening. These include zebrafish, Drosophila (fruit flies), and the nematode C. elegans (roundworms) (Katner et al., 2019; Krook et al., 2019; Mocelin et al., 2019; Ogelade et al., 2015; Pardo-Martin et al., 2013). These models preserve the complexity and architecture of intact organs and have proven to be useful in studying alcohol’s mechanisms underlying withdrawal, tolerance, and dependence, as well as the neuromolecular and circuit basis for alcohol-seeking behaviors (e.g., craving) and alcohol-induced neuroimmune dysfunction (Kaun et al., 2012; Mocelin et al., 2019; Petruccelli and Kaun, 2019).

In summary, understanding how preclinical and clinical screening models can be linked to clinical efficacy and safety is vital for improving the accuracy and efficiency of the drug development pipeline. In particular, more attention is needed to determine how preclinical models can be used to inform the design and selection of individuals in clinical trials. This is essential if we want to improve our ability to decrease the time needed for drug development, improve the success rate for these drugs, and make the drug development process less expensive.

ENGAGE PHARMACEUTICAL/BIOTECHNICAL COMPANIES IN MEDICATIONS DEVELOPMENT FOR AUD

The time, costs, and risks involved in the drug development process can be at least somewhat mitigated through collaboration—with the government, academia, and pharmaceutical/biotechnology companies all playing key roles. For example, pharmaceutical companies have the financial resources and expertise to invest in a solid medication-development program, including expensive Phase 3 trials and extensive marketing campaigns (Litten et al., 2014). Importantly, companies are uniquely disposed to possess a library of untested novel compounds (and those that have failed testing for other indications). Of course, companies themselves could evaluate these compounds for an AUD indication. However, just as important, they could make the compounds available through licensing agreements to other companies and academic researchers who otherwise would have a difficult time obtaining novel compounds. Likewise, pharmaceutical companies stand to gain by collaborating with government/academia as the industry relies heavily on academia to identify or refine novel molecular targets for study. Academia and government can also conduct studies, advise on the development of animal models and the human laboratory paradigms needed for screening promising compounds and determining their mechanism of action, and provide expertise on designing and conducting alcohol clinical trials.

Over the past decade, the pharmaceutical industry has shown greater interest in developing medications for treating AUD, as evidenced by 3 recent clinical trials: ABT-436, a vasopressin V1b antagonist (Ryan et al., 2017); LY2940094, a nociception receptor antagonist (Post et al., 2016); and LY2196044, an opioid receptor antagonist (Wong et al, 2014). NIAAA continues to seek ways to facilitate pharmaceutical/biotechnical industry involvement. As described above, NIAAA has spearheaded efforts to standardize the use of animal models for screening candidate medications. To date, more than a dozen companies have used these models to test novel compounds. As noted above, NIAAA also helped to launch a human laboratory model, which has been used at multiple sites to examine varenicline. A second study is currently underway to examine the compound ANS-6637. In addition to the screening models, NIAAA initiated a Small Business Innovation Research (SBIR) and Small Business Technology Transfer (STTR) to assist with the IND-enabling development and human Phase 1 studies of promising compounds (Litten et al., 2016b). NIAAA previously issued a similar program for academic researchers who were interested in developing novel compounds. NIAAA created a network of sites for conducting Phase 2 clinical trials geared toward speeding the trial time, so that studies are completed within 1.5 years—which is also proving to be attractive to pharmaceutical partners (Litten et al., 2016b). These sites follow the requirements set forth under the Good Clinical Practice (GCP) guidelines. In the last 10 years, 5 clinical trials have been completed and the results were published in peer-reviewed journals. Three of these trials were conducted in collaboration with pharmaceutical companies (Falk et al., 2019b; Litten et al., 2012; Ryan et al., 2017). The idea of facilitating partnership among government institutions, pharmaceutical companies, and academia also is the backbone of the recently created National Center for Advancing Translational Sciences (NCATS) at the NIH. NCATS-led initiatives range across the entire biomedical field and include major programs for repurposing new compounds developed by companies for new indications. An example pertinent to the AUD field is the ongoing effort to test an inverse agonist of the ghrelin receptor under a NCATS program, which is facilitated by an ongoing collaboration between the NIAAA/NIDA intramural research programs, the University of Rhode Island, and a pharmaceutical company (Lee et al., 2018; NCT: 02707055).

Other NIAAA efforts designed to improve clinical trial efficacy have centered on providing research data/analyses to the FDA to aid in developing evidence-based guidelines for pharmaceutical companies to conduct pivotal clinical trials. As a result, the FDA issued guidance for developing drugs for the treatment of AUD, the first guidance in the field of addiction (https://www.fda.gov/media/91222/download). These guidelines are particularly important for establishing alcohol consumption endpoints that truly reflect recovery-related patient improvements in feeling and functioning. NIAAA has been part of the Alcohol Clinical Trials Initiative (ACTIVE) whose purpose is to improve alcohol clinical trial methodology. The group also consists of academic researchers, representatives from the FDA and European Medicines Agency (EMA), and a number of pharmaceutical companies. The ACTIVE group conducted secondary data analyses on a variety of clinical trials and epidemiological databases to arrive at and validate a new primary endpoint for AUD, the “percent subjects with no heavy drinking” (Falk et al., 2010). This endpoint, in addition to abstinence, is now accepted by the FDA as a primary endpoint for Phase 3 trials. The group is now working on examining another endpoint—a “reduction in the World Health Organization’s (WHO’s) risk drinking levels,” an endpoint that has been accepted by EMA in alcohol clinical trials, but not the FDA (https://www.ema.europa.eu/en/documents/scientific-guideline/guideline-development-medicinal-products-treatment-alcohol-dependence_en.pdf). This harm reduction endpoint appears to be at least as sensitive as the abstinence and “percent subjects with no heavy drinking” endpoints and captures more positive responses when compared with the 2 existing endpoints (Falk et al., 2019a; Hasin et al., 2017; Knox et al, 2019, 2018; Witkiewitz et al., 2019b, 2017, 2018).

Finally, NIAAA is working to inform the industry of the value in developing and marketing a medication for AUD. One such benefit is the fact that very little competition exists in the marketplace for AUD medications. Only one medication is being actively promoted in the United States, Vivitrol^®, a long-acting injectable of naltrexone marketed by Alkermes. Yet, the vast majority of the 15 million Americans diagnosed with AUD each year are untreated, by medications or any other therapeutics, and many more are experiencing problematic drinking who do not meet the criteria of AUD. Clearly, there are financial incentives for the private industry to become involved in developing medications for AUD and room for additional public–private networking and collaborations in this area.

FACILITATE THE USE OF ALCOHOL TREATMENT MEDICATIONS IN CLINICAL PRACTICE

Clinical trials provide evidence that several medications can be effective in treating AUD (Litten et al., 2018). But like other complex diseases, such as depression, which has a similar prevalence, one medication does not work for all patients. Antidepressants and AUD medications share similar efficacies in clinical trials; yet unlike antidepressants, alcohol medications are prescribed at a fraction of the rate. (Litten, 2016; Litten et al., 2016b). It is estimated that less than 4 percent of AUD patients are prescribed FDA-approved alcohol treatment medications (Litten, 2016; Litten et al., 2016b), whereas antidepressant use has continued to climb, from 7.6 percent in 1999–2002 to 12.7 percent in 2011–2014 (https://www.cdc.gov/nchs/products/databriefs/db283.htm). Beyond addiction and mental health disorders, another compelling example is that naltrexone’s and acamprosate’s effects on preventing relapse to heavy drinking or any drinking, respectively, present with a number needed to treat (NNT) of 12 in both cases, which are similar to that of beta-blockers’ effects in preventing hospitalization and death in heart failure; still, beta-blockers are often referred to as “highly effective,” whereas naltrexone and acamprosate are often labeled as modestly effective medications (Heilig and Leggio, 2016; Shibata et al., 2001)).

The reasons why alcohol treatment medications are under-prescribed in clinical practice in the United States have been well documented. Clinicians and patients may not be aware that medications exist or may be skeptical that they may work; treatment may not be covered under some insurance plans; specialty treatment programs may lack clinicians authorized to prescribe medications; and treatment staff may be reluctant to use “drugs” in the treatment of substance abuse (Ducharme et al., 2006; Knudsen and Roman, 2014; Mark et al., 2003; Thomas et al., 2003) because of the false concept that it is just replacing a drug with another. Williams and colleagues (2018) reported barriers to medications use in primary care settings based on findings from 5 VA clinics. They found clinicians often had gaps in their knowledge of and experience with AUD and did not believe AUD medications could take the place of specialty treatment. Staff also had stigmatizing, prejudicial attitudes toward patients with alcohol problems, often believing these patients could not be “cured” because they would choose addiction over treatment. Patients also report encountering bias when receiving alcohol treatment (Livingston et al, 2011; Wakeman and Rich, 2018). Patients have expressed concerns about consequences of disclosing alcohol problems, confidentiality issues, are reluctant to acknowledge a problem, and have difficulty accessing addiction treatment; and systems-level factors among clinicians include time pressure, resources, and lack of space, which may all represent additional barriers against providing effective treatments (McNeely et al., 2018). In addition to these barriers, addiction medicine often is not included in medical education curriculum, or is only tangentially addressed, thus contributing to clinicians’ lack of awareness of the problems caused by AUD. Although less than 8 percent of patients with AUD sought treatment for their disorder in the past year, studies show that they often do seek care from primary care physicians, not for their drinking but for alcohol-related medical problems (O’Connor et al., 2011; Rehm et al., 2016). Thus, it is likely that many primary care physicians are missing an opportunity to treat AUD patients. Reducing the stigma and moving toward a medical and public health approach to address AUD is not only important to further increase the range of acceptable treatment options but to execute the translation of promising developments in addictions treatment into everyday clinical practice and public health decision making.

NIAAA, other Federal agencies, and professional organizations are working to mitigate the barriers to treatment and to promote the use of medications, although more work is needed in these areas. To help fill knowledge and training gaps, for example, NIAAA published Helping Patients Who Drink Too Much: A Clinician’s Guide (https://www.niaaa.nih.gov/guide). This guide and related online continuing education courses give guidance and training to clinicians on how to screen for heavy drinking, conduct brief interventions, offer medications, and provide follow-up support. In partnership with the Substance Abuse and Mental Health Administration, NIAAA developed a guide, Medications for the Treatment of Alcohol Dependence, Other Alcohol Use Disorders, and Related Comorbidities: A Brief Guide, to give basic information on diagnosing and treating patients with AUD. The American Society of Addiction Medicine offers many in-person continuing education events at professional meetings as well as an extensive e-Learning Center, which includes brief and in-depth training courses on addiction medicine for both primary care and specialist providers. The Association for Multidisciplinary Education and Research in Substance Use and Addiction is working to raise awareness among its members on the use of medications for AUD and other substance use disorders. One of its goals is to educate internists to increase their comfort level in prescribing these medications.

NIAAA also is working to give clinicians fundamental information on alcohol and health that can be applied across all areas of health care—from primary care to specialty treatment centers. This initiative will make it easier for healthcare providers to access information on alcohol, including guidance on how to use AUD medications to help their patients. It will also focus on recognizing, confronting, and overcoming the stigma associated with AUD to provide the very best care possible to patients with alcohol problems.

Another new NIAAA resource designed to facilitate treatment is the Alcohol Treatment Navigator (https://alcoholtreatment.niaaa.nih.gov/). The Navigator gives healthcare professionals a way to locate (1) board-certified addiction medicine physicians who can provide external prescribing support and (2) specialist therapists who can follow up with patients and deliver evidence-based behavioral therapy. NIAAA has recently issued a new Navigator portal for healthcare professionals (https://alcoholtreatment.niaaa.nih.gov/healthcare-professionals).

We recognize that these resources can themselves present a barrier. Clinicians receive a barrage of information on topics related to improving practice. It can be a challenge to know where to start and how to incorporate AUD medications into a busy practice with limited resources. The best approaches will use a “multichannel” approach to engage clinicians and help them gain the skills and information they need to make better use of today’s medications. These channels range from presentations at professional meetings and the use of “influencers” in social media to partnerships among government and professional organizations.

At the systems level, we need to look at strengthening medical school curricula in addiction medicine (Wood et al., 2013). In a major national effort to address this issue, the American College of Academic Addiction Medicine is developing and supporting graduate medical education fellowships and working to integrate addiction medicine into medical schools and residencies. Other system-level strategies include working with healthcare organizations to revise and standardize clinical practice guidelines and quality standards related to alcohol screening, intervention, and treatment. NIAAA clinician’s guides, for example, have been adopted into clinical practice guidelines of large healthcare organizations. System-level efforts are now underway at these large healthcare organizations (Sterling et al., 2019) and at the Center for Disease Control and Prevention, with new initiatives to incorporate alcohol screening and brief intervention workflows into electronic health record systems.

Everyone who goes to a clinic or hospital receives as standard practice a basic check that includes blood pressure, heart rate, and weight. This is routinely done, no matter how often the individual goes to a clinical setting. Risky alcohol use remains a leading cause of mortality and morbidity worldwide, yet AUD screening and treatment, including the prescription of medications, has yet to become a routine part of clinical practice. During the next few years, as the menu of medication options expands, we need to give special attention to building an infrastructure for alcohol treatment in general, and the use of medications in particular, and to make healthcare providers and consumers aware of all the evidence-based treatments available to them.

SUMMARY

During the past 25 years, advances have been made in developing medications to treat AUD, but more work remains to be done. We need to improve our ability to identify targets that yield more effective candidate compounds, advance personalized medicine, develop screening models to predict a high probability of clinical success, engage pharmaceutical companies so that they will consider developing medications to treat AUD, and facilitate the use of alcohol treatment medications in clinical practice. New developments in tools, technologies, and methodologies offer many exciting possibilities for advancing these areas. It is clear, however, given the complexities of bringing a drug to market, that AUD medications development cannot take place in a vacuum, or in one lab. This process requires scientists from the alcohol-research field, as well as scientists spanning a full spectrum of related fields—from basic science to information technology and health services research. It will require close collaboration with pharmaceutical and biotech companies, with academic researchers, with government agencies, such as NIAAA and the FDA, as well as with public and private insurers, program and systems administrators, healthcare organizations, and the patients themselves. With the mobilization of such teams, we will be able to make the greatest strides in medications development, and in providing relief for patients and their families.