Connectome-based predictive modeling of concurrent and longitudinal substance use vulnerability in adolescence

Area of Science:

• Neuroscience
• Developmental Psychology
• Public Health

Background:

• Adolescent substance use is a major public health concern.
• Understanding its neural underpinnings is crucial for prevention and treatment.
• Existing brain network features struggle to capture dynamic and multifaceted aspects of substance use.

Purpose of the Study:

• To model adolescent substance use facets (intent, access, family history) over time.
• To investigate the predictive power of traditional and emergent brain connectivity metrics.
• To examine developmental shifts in brain-behavior relationships related to substance use risk.

Main Methods:

• Utilized Baseline and 2-Year Follow-Up data from the ABCD Bids Community Collection (ABCC).
• Operationalized substance use into intent, access, and family-developmental history.
• Employed traditional and connectome embedding metrics for resting-state connectivity analysis.
• Trained predictive models for each substance use facet at multiple timepoints.

Main Results:

• Predictive accuracy for substance use facets increased significantly from baseline to follow-up.
• Intent to use was most predictable at baseline; access and family history improved prediction at follow-up.
• Both traditional and emergent connectivity metrics showed similar predictive performance.

Conclusions:

• Neurobiological correlates of adolescent substance use are dynamic and change across development.
• Distinct substance use facets and developmental timing are critical for understanding risk.
• Early individual differences in brain connectivity predict later substance use vulnerabilities.