Who is alcohol cue-reactive? A machine learning approach.

Area of Science:

• Neuroscience
• Psychiatry
• Data Science

Background:

• Alcohol cue-exposure is vital in alcohol use disorder (AUD) research.
• Significant variability exists in AUD patients' alcohol cue-reactivity.
• Identifying predictors of this heterogeneity is crucial for targeted interventions.

Purpose of the Study:

• To apply machine learning models to identify clinical and sociodemographic predictors of subjective alcohol cue-reactivity (ALCUrge).
• To compare the predictive accuracy of different machine learning models in this context.

Main Methods:

• 139 individuals with AUD underwent an alcohol cue-exposure paradigm and provided clinical/sociodemographic data.
• Subjective alcohol urge (ALCUrge) was measured using the Alcohol Urge Questionnaire post-exposure.
• Lasso regression, Ridge regression, and Random Forest models were employed to identify predictors.

Main Results:

• Lasso regression demonstrated the highest predictive accuracy (RMSE = 9.48), outperforming Random Forest and Ridge regression.
• Top predictors of ALCUrge included pre-cue alcohol urge, compulsive alcohol behaviors, tonic craving, smoking status, and biological sex.
• Higher pre-cue urge, compulsive behaviors, tonic craving, and smoking predicted greater ALCUrge, while being female predicted lower ALCUrge.

Conclusions:

• The study clarifies the overlap between compulsive alcohol-related behaviors and cue-induced urges.
• Biological sex and cigarette smoking emerged as significant factors influencing alcohol cue-reactivity variability.
• Findings enhance understanding of AUD heterogeneity and inform personalized treatment strategies.