Neurobiological subtypes in alcohol use disorder and their phenotypic and clinical profiles

Alcohol Use Disorder (AUD) is associated with vast clinical, behavioral, and neurobiological heterogeneity that hinder both the efficacy of current interventions and future treatment development. While recent efforts have focused on stratifying psychiatric disorders based on underlying neurobiological markers, such approaches remain largely unexplored in AUD. To address this, we empirically-derived neurobiological subtypes of AUD and evaluated their reproducibility and clinical relevance. Participants with complete resting-state functional MRI (rs-fMRI), phenotypic and clinical data from the Human Connectome Project were included (N = 668; 58% Female; 22% AUD [mild AUD: n = 109; moderate-severe AUD: n = 38], 40% Female within AUD). Neurobiological subtypes of AUD were identified using a semi-supervised clustering approach based on graph theory metrics derived from whole-brain rs-fMRI. Subtypes were extensively validated for reproducibility using resampling, permutation testing, and temporal stability, and for clinical relevance using comprehensive assessments of multi-modal behaviors and clinical characteristics. Two robust and distinct neurobiological subtypes of AUD with unique clinical profiles emerged. Subtype 1 showed greater sensory-motor integration, but lower frontoparietal/salience integration, with elevated externalizing behaviors. Subtype 2 demonstrated the opposite pattern: greater integration in frontoparietal, salience, and default-mode networks, and lower sensory-motor integration, characterized primarily by internalizing behaviors. Subtypes did not differ by AUD symptoms or use characteristics. Importantly, the subtype structure and phenotypic profiles remained robust when including AUD individuals with comorbid cannabis dependence, supporting generalizability and translational relevance to clinically comorbid populations. In summary, we identified neurobiological subtypes of AUD with distinct phenotypic and clinical profiles, capturing greater nuance than previous phenotypically-derived subgroups, while remaining interpretable and clinically meaningful. These findings provide a robust framework for understanding AUD's underlying mechanisms and underscore their potential for a neurobiologically informed approach to nosology, treatment allocation, and personalized interventions in AUD.