This issue of the
Journal includes an important article by King and colleagues (
1) in which subjective responses to alcohol are longitudinally tested in the laboratory and related to drinking outcomes 10 years later. Supported by state-of-the-art methods and outstanding retention rates, the study found that individuals who reported greatest alcohol stimulation, liking, and wanting at the initial alcohol challenge in the laboratory were most likely to develop an alcohol use disorder (AUD) 10 years later. Importantly, the rewarding subjective effects of alcohol (i.e., stimulation and euphoria) did not wane as AUD symptoms developed. This is contrary to the hypothesis that as AUD develops and progresses, there is a shift away from the positive reinforcing effects of alcohol and toward negative reinforcement marked by alleviation of protracted withdrawal and negative mood (
2,
3). This finding is in turn consistent with predictions from the incentive-sensitization model of addiction whereby alcohol “wanting” increased across a decade in individuals who developed AUD (
4).
Before discussing the implication of these findings to understanding AUD vulnerability in humans, it is critical to fully appreciate how challenging, yet necessary, the translation of preclinical theories of addiction to human and clinical samples really is (
5). It is also critical to note that these robust findings provide unique opportunities for reverse translation, whereby discoveries at the human level of analysis can inform preclinical models (
6). Within the timely framework of translational research in AUD, and addiction broadly, it is our goal to dissect the seminal findings of King et al. toward a nuanced understanding that can provide concrete avenues for clinical and preclinical research.
We begin with a discussion of the sample itself and how sample characteristics can contextualize the findings and implications. As seen in Table 1 in the article by King et al., this is, on average, an educated and predominantly White sample. There are lower levels of nicotine and substance use disorder in the sample than in the general population, particularly in samples with AUD where comorbid substance use is common. Levels of depressive symptomatology are also noticeably below clinical thresholds. Even among those with AUD at year 10, the average number of heavy drinking days in the past month was 14 days, and the average number of AUD symptoms was 4.5 out of 11 possible symptoms. These characteristics suggest that while highly informative, this sample may not represent a large proportion of individuals who develop AUD in the context of lower socioeconomic level and educational achievement at baseline. In the study by King et al., the progression to AUD symptoms is also marked by drinking levels that are arguably on the lower range of alcohol consumption in AUD samples, particularly treatment-seeking samples (
7). In other words, the negative consequences associated with alcohol use in this sample may not be sufficiently demonstrated to represent a broader spectrum of AUD and addiction in our communities. The sample clearly maintained a positive hedonic response to alcohol consistent with the incentive-sensitization model, as concluded by the authors. However, the notion that these findings can effectively refute an alternative pathway to AUD, namely through allostatic changes, may be premature. Specifically, it may be that higher levels of AUD severity (at follow-up) and perhaps vulnerability to AUD (at baseline) are necessary to unmask the “dark side of addiction” wherein the positive hedonic effects of alcohol are no longer a primary driver of alcohol intake, which is instead driven by a negative emotional state and withdrawal. In sum, although this study confirms the incentive-sensitization model for some drinkers, it does not properly refute other models, primarily the allostatic model.
From a translational point of view, what do these influential findings suggest as next steps in AUD clinical research? First, we need samples that represent a wider range of vulnerabilities at baseline, including genetic and environmental risk factors, such as lower socioeconomic status and higher family history density for addiction. This point is illustrated by the fact that at baseline, the average age of the sample was approximately 25 years, and the sample comprised heavy drinkers without significant alcohol pathology. Clinically, age 25 is past the age of risk for AUD and addiction development in many communities (
8). Thus, it can be argued that these individuals may in fact display some resilience by not having developed AUD by age 25. Second, this study argues for a wider inclusion of participants in alcohol administration studies. Given the associations between socioeconomic status and AUD, with individuals with lower socioeconomic status experiencing more alcohol-related consequences (
9), it would be imperative to include participants from lower socioeconomic levels to understand whether the results found as part of King et al.’s study are generalizable to higher-risk populations. The increased prevalence of alcohol-related consequences in racial and ethnic minority populations warrants inclusion of greater proportions of participants from racial and ethnic backgrounds in human laboratory studies so that the populations represented in such studies align with the epidemiological findings for those most at risk for an AUD (
10). Lastly, it is understandable that given the minimum legal drinking age of 21 years in the United States, the authors were restricted to that as the lower age limit. There has been consistent debate in the field about the ethics of alcohol administration and the need to balance these constraints with the critical gains of understanding AUD development in an ecologically valid fashion. Ecological momentary assessment methods offer one approach to studying alcohol consumption in younger populations, albeit with a loss of the rigorous experimental control used in the King et al. study. In brief, translational research in AUD would benefit greatly from studying the individuals most vulnerable and most likely to end up in treatment settings.
From a reverse translational point of view, what do these findings suggest as next steps in AUD preclinical research? An important direction in translating such findings consists of using genetically diverse mouse models (
11), which in turn may be useful in identifying not only “punishment-resistant” alcohol intake, where rodents continue consuming alcohol despite aversive punishments such as electric shock, but also reward-driven intake. Further, when considering the translation of constructs that may be inherently subjective, as the ones addressed in King et al., it is critical to consider the argument that language itself may limit their translatability (
12). Preclinical models rely on learning—where an animal must have repeated direct experiences with a drug, context, and environment—to experience what humans can communicate through language. This issue is particularly critical in King et al.’s study, as subjective effects are inherently tied to human language. While these issues pose additional barriers to translation, the overarching recommendation is for the development of animal models that can effectively represent reward-driven alcohol use because the clinical literature, including this hallmark study by King and colleagues, suggests that reward-driven alcohol use constitutes a critical component of AUD.
In sum, the study by King et al. (
1) provides a nuanced representation of AUD heterogeneity with a compelling argument for the incentive-salience model, whereby “wanting” of alcohol and reward-driven alcohol use remain critical drivers in AUD phenomenology. While not ruling out competing models that emphasize a transition to negative reinforcement, this study clearly makes the argument for reward-blocking as an ongoing treatment target (
13). A balanced interpretation of these results in the broader context of AUD risk and resilience suggests that although reward-driven alcohol use remains prominent as an etiological risk marker and treatment target, efforts to capture negative-reinforcement contributions to AUD should include diverse populations with greater clinical severity and less social capital. In closing, we commend King and colleagues for a critical study bridging behavioral pharmacology–driven biomarkers into AUD development and maintenance over the span of 10 years in adult development. This impressive undertaking is sure to pave the way for translational research in the field, which is much needed to inform the prevention and treatment of AUD.