Varenicline has shown promise for treating alcohol use disorder (AUD); however, not everyone will respond to varenicline. Machine-learning methods are well suited to identify treatment responders. In the present study, we examined data from the National Institute on Alcohol Abuse and Alcoholism Clinical Intervention Group multisite clinical trial of varenicline using two machine-learning methods. Baseline characteristics taken from a randomized clinical trial of varenicline were examined as potential moderators of treatment response using qualitative interaction trees (N = 199) and group least absolute shrinkage and selection operator interaction nets (N = 200). Results align with prior research, highlighting smoking status, AUD severity, medication adherence, and drinking goal as predictors of treatment response. Novel findings included the interaction between age and cardiovascular health in predicting clinical response and stronger medication effects among individuals with lower craving. With increased integration of machine-learning methods, studies that effectively integrate methods and medication development have high potential to inform clinical practice.

RATIONALE: Alcohol administration and cue-reactivity paradigms are frequently used to screen for the initial efficacy of medications for alcohol use disorder (AUD). While medication effects on the primary outcomes for these paradigms are assumed to be qualitatively related, there is a critical lack of quantitative evidence to support this hypothesis. OBJECTIVES: The study aims to test the relationship between medication effect sizes on subjective response to alcohol administration and medication effect sizes for cue-induced craving to cue exposure, using meta-analysis. METHODS: Systematic literature searches were conducted to identify randomized trials, wherein AUD medications were tested using the alcohol administration and/or cue-reactivity paradigms. From these studies, descriptive statistics were collected to compute medication effect sizes on the primary outcomes for each respective paradigm. With medication as the unit of analysis, medication effect sizes in alcohol administration studies were compared with medication effect sizes in cue-reactivity studies using the Williamson-York regression which allows for meta-regression across independent samples. RESULTS: Medication effect sizes on alcohol-induced stimulation and alcohol-induced craving were not significantly associated with medication effect sizes on cue-induced alcohol craving (k stimulation = 10 medications, [Formula: see text] and k craving = 11 medications, [Formula: see text] (SE = 0.237), [Formula: see text]), respectively. Medication effect sizes on alcohol-induced sedation were significantly associated with medication effects on cue-induced craving (k = 10 medications, [Formula: see text] (SE = 0.258), [Formula: see text]), such that medications that increased alcohol-induced sedation were more likely to reduce cue-induced alcohol craving. CONCLUSIONS: With the exception of alcohol-induced sedation, there is little quantitative evidence of medication effects on subjective response domains measured during alcohol administration parallel medication effects on cue-induced alcohol craving. To provide additional context to the current study, future work should examine whether cue-reactivity findings predict clinical trial outcomes.

Short Summary: The Letter to the Editors regarding our article was reviewed. The take-home message is that substantively, the authors of the letter are referencing a paper that asks a different research question in a different set of studies. When we ask different questions, we are not surprised when we reach different answers.

Aims

To test whether two critical design features, inclusion criteria of required pre-trial abstinence and pre-trial alcohol use disorder (AUD) diagnosis, predict the likelihood of detecting treatment effects in AUD pharmacotherapy trials.

Methods

This secondary data analysis used data collected from a literature review to identify randomized controlled pharmacotherapy trials for AUD. Treatment outcomes were selected into abstinence and no heavy drinking. Target effect sizes were calculated for each outcome and a meta-regression was conducted to test the effects of required pre-trial abstinence, required pre-trial AUD diagnosis, and their interaction on effect sizes. A sub-analysis was conducted on trials, which included FDA-approved medications for AUD.

Results

In total, 118 studies testing 19 medications representing 21,032 treated participants were included in the meta-regression analysis. There was no significant effect of either predictor on abstinence or no heavy drinking outcomes in the full analysis or in the sub-study of FDA-approved medications.

Conclusion

By examining these design features in a quantitative, rather than qualitative, fashion the present study advances the literature and shows that requiring AUD diagnosis or requiring pre-trial abstinence do not impact the likelihood of a significant medication effect in the trial.

RATIONALE: The alcohol cue exposure paradigm is a common method for evaluating new treatments for alcohol use disorder (AUD); however, it is unclear if medication-related reductions in cue-induced craving in the human laboratory can predict the clinical success of those medications in reducing alcohol consumption during clinical trials. OBJECTIVES: To use a novel meta-analytic approach to test whether medication effect sizes on cue-induced alcohol craving are associated with clinical efficacy in clinical trials. METHOD: We searched the literature for medications tested for AUD treatment using both the alcohol cue-reactivity paradigm and randomized clinical trials (RCTs). For alcohol cue-reactivity studies, we computed medication effect sizes for cue-induced alcohol craving (k = 36 studies, 15 medications). For RCTs, we calculated medication effect sizes for heavy drinking and abstinence (k = 139 studies, 19 medications). Using medication as the unit of analysis, we applied the Williamson-York bivariate weighted least squares estimation to account for errors in both independent and dependent variables. We also conducted leave-one-out cross validation simulations to examine the predictive utility of cue-craving medication effect sizes on RCT heavy drinking and abstinence endpoints. RESULTS: There was no significant relationship between medication effects on cue-induced alcohol craving in the human laboratory and medication effects on heavy drinking ( β ^ = 0.253, SE = 0.189, p = 0.090) and abstinence ( β ^ = 0.829, SE = 0.747, p = 0.133) in RCTs. CONCLUSIONS: The preliminary results of the current study challenge the assumption that alcohol cue-reactivity alone can be used as an early efficacy indicator for AUD pharmacotherapy development. These findings suggest that a wider range of early efficacy indicators and experimental paradigms be considered for Phase II testing of novel compounds.

Background

Human laboratory paradigms are a pillar in medication development for alcohol use disorders (AUD). Neuroimaging paradigms, in which individuals are exposed to cues that elicit neural correlates of alcohol craving (e.g., mesocorticolimbic activation), are increasingly utilized to test the effects of AUD medications. Elucidation of the translational effects of these neuroimaging paradigms on human laboratory paradigms, such as self-administration, is warranted. The current study is a secondary analysis examining whether alcohol cue-induced activation in the ventral striatum is predictive of subsequent alcohol self-administration in the laboratory.

Methods

Non-treatment-seeking heavy drinkers of East Asian descent (n = 41) completed a randomized, placebo-controlled, double-blind, crossover experiment on the effects of naltrexone on neuroimaging and human laboratory paradigms. Participants completed 5 days of study medication (or placebo); on day 4, they completed a neuroimaging alcohol taste cue-reactivity task. On the following day (day 5), participants completed a 60-minute alcohol self-administration paradigm.

Results

Multilevel Cox regressions indicated a significant effect of taste cue-elicited ventral striatum activation on latency to first drink, Wald χ²  = 2.88, p = 0.05, such that those with higher ventral striatum activation exhibited shorter latencies to consume their first drink. Similarly, ventral striatum activation was positively associated with total number of drinks consumed, F(1, 38) = 5.90, p = 0.02. These effects were significant after controlling for alcohol use severity, OPRM1 genotype, and medication. Other potential regions of interest (anterior cingulate, thalamus) were not predictive of self-administration outcomes.

Conclusions

Neuroimaging alcohol taste cue paradigms may be predictive of laboratory paradigms such as self-administration. Elucidation of the relationships among different paradigms will inform how these paradigms may be used synergistically in experimental medicine and medication development.

Background

Human laboratory paradigms are a pillar in medication development for alcohol use disorders (AUD). Neuroimaging paradigms, in which individuals are exposed to cues that elicit neural correlates of alcohol craving (e.g., mesocorticolimbic activation), are increasingly utilized to test the effects of AUD medications. Elucidation of the translational effects of these neuroimaging paradigms on human laboratory paradigms, such as self-administration, is warranted. The current study is a secondary analysis examining whether alcohol cue-induced activation in the ventral striatum is predictive of subsequent alcohol self-administration in the laboratory.

Methods

Non-treatment-seeking heavy drinkers of East Asian descent (n = 41) completed a randomized, placebo-controlled, double-blind, crossover experiment on the effects of naltrexone on neuroimaging and human laboratory paradigms. Participants completed 5 days of study medication (or placebo); on day 4, they completed a neuroimaging alcohol taste cue-reactivity task. On the following day (day 5), participants completed a 60-minute alcohol self-administration paradigm.

Results

Multilevel Cox regressions indicated a significant effect of taste cue-elicited ventral striatum activation on latency to first drink, Wald χ²  = 2.88, p = 0.05, such that those with higher ventral striatum activation exhibited shorter latencies to consume their first drink. Similarly, ventral striatum activation was positively associated with total number of drinks consumed, F(1, 38) = 5.90, p = 0.02. These effects were significant after controlling for alcohol use severity, OPRM1 genotype, and medication. Other potential regions of interest (anterior cingulate, thalamus) were not predictive of self-administration outcomes.

Conclusions

Neuroimaging alcohol taste cue paradigms may be predictive of laboratory paradigms such as self-administration. Elucidation of the relationships among different paradigms will inform how these paradigms may be used synergistically in experimental medicine and medication development.

Aims

Natural processes of change have been documented in treatment-seekers who begin to reduce their drinking in anticipation of treatment. The study examined whether non-treatment-seeking problem drinkers would engage in drinking reduction in anticipation of participating in a research study.

Methods

Non-treatment-seeking problem drinkers (n = 935) were culled from five behavioral pharmacology studies. Participants reported on their alcohol use during the past 30 days using the Timeline Followback. Cluster analysis identified distinct groups/clusters based on drinking patterns over the 30-day pre-visit period. The identified clusters were compared on demographic and clinical measures.

Results

Three distinct clusters were identified (a) heavy-decreasing drinking group (n = 255, 27.27%); (b) a moderate-stable drinking group (n = 353, 37.75%) and (c) low-stable drinking group (n = 327, 34.97%). The three clusters differed significantly on a host of measures including pre-visit drinking (age at first drink, drinking days, drinks per week, drinks per drinking day), alcohol use severity, alcohol craving, readiness for change, depression and anxiety levels. These differences were alcohol dose-dependent such that the heavier drinking group reported the highest levels on all constructs, followed by the moderate group, and the low drinking group last.

Conclusions

Baseline drinking patterns of non-treatment-seekers were generally stable and pre-visit reductions were only observed among the heavy drinking group. This generally stable pattern stands in contrast to previous reports for treatment-seeking samples. Nevertheless, the heavier drinking group, which is most similar to treatment-seekers, displayed pre-study drinking reduction. Overall, naturalistic processes of change may pose less of a threat to randomization and testing in this population.