An integrated multimodal model of alcohol use disorder generated by data-driven causal discovery analysis

Rawls, Eric; Kummerfeld, Erich; Zilverstand, Anna · 2021 · Crossref

DOI: 10.1038/s42003-021-01955-z

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the complexity of alcohol use disorder (AUD) by investigating the multifactorial neurobehavioral mechanisms that drive AUD severity. While early addiction theories focused on single mechanisms, current models recognize multiple contributing domains but often rely on expert consensus rather than data-driven evidence. The authors aimed to generate an integrated, multimodal causal model of AUD by analyzing the relationships between brain connectivity, phenotypic factors, and symptom severity, thereby identifying specific causal pathways that could inform targeted treatments. The researchers utilized data from the Human Connectome Project (HCP), comprising 926 participants (54% female), of whom 22% met criteria for AUD. They first applied exploratory factor analysis to reduce 100 phenotypic measures into 18 underlying domains, such as fluid cognition, social support, and negative affect. Additionally, they extracted functional connectivity metrics from 12 resting-state brain networks. To determine causal relationships, the authors employed Greedy Fast Causal Inference (GFCI), a machine learning technique capable of handling unmeasured confounders in observational data. This analysis generated a partial ancestral graph linking brain connectivity, phenotypic domains, and AUD symptom counts, which was further validated using structural equation modeling. The results revealed a hierarchical causal structure where influence propagates from brain connectivity to cognition, then to social factors, and finally to affective/psychiatric functions before impacting AUD severity. Specifically, connectivity in the ventral attention/language network influenced default mode and other self-reflective networks. Frontoparietal connectivity was linked to fluid cognition, which subsequently influenced crystallized IQ, working memory, and language abilities. These cognitive factors causally influenced social domains, particularly agreeableness and social support. Low social support and agreeableness mediated the effect of cognition on negative affect and externalizing symptoms. Crucially, externalizing psychopathology (aggression, rule-breaking) was found to fully mediate the impact of all other measured factors on AUD severity. The model also confirmed a direct causal link between negative affect and AUD, mediated through conscientiousness and attention. The significance of this work lies in its data-driven expansion of addiction models. The findings underscore the critical role of social factors, such as social support and agreeableness, as mediators between cognitive deficits and psychiatric symptoms in AUD, a relationship often overlooked in neurobiological models. Furthermore, the study highlights fluid cognition and crystallized IQ as upstream causal factors in the hierarchy of AUD influences, suggesting that cognitive interventions may be relevant. By demonstrating that externalizing symptoms mediate the effects of broader neurobehavioral factors, the paper provides empirical evidence for the multifactorial nature of AUD and identifies specific targets for future therapeutic development.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• polydrug