Brain Imaging Identified a Neural Network That Predicts Problem Cannabis Use Across Three Independent Samples

Machine learning identified a brain connectivity pattern during reward processing that predicted problem cannabis use in college students, generalized to European adolescents, and was linked to worse treatment outcomes in adults with cannabis use disorder.

Lichenstein, Sarah D et al.·Biological psychiatry·2025·Moderate EvidenceObservational
RTHC-06948ObservationalModerate Evidence2025RETHINKTHC RESEARCH DATABASErethinkthc.com/research

Quick Facts

Study Type
Observational
Evidence
Moderate Evidence
Sample
N=33

What This Study Found

A whole-brain machine learning approach identified a "problem cannabis risk network" from reward task brain connectivity data in college students. This network generalized to predict cannabis use in 1,320 European adolescents and was linked to higher addiction severity and poorer treatment outcomes in 33 treatment-seeking adults. The network was specific to cannabis and did not predict alcohol use outcomes.

Key Numbers

Discovery sample: 191 college students (58% female). Validation sample 1: 1,320 European adolescents (53% female). Clinical sample: 33 treatment-seeking adults (9% female). The network was specific for cannabis vs alcohol across all 3 datasets.

How They Did This

Data-driven machine learning analysis of reward task functional connectivity in 191 college students (58% female). External validation in 1,320 European adolescents/emerging adults from the IMAGEN study and 33 adults seeking treatment for cannabis use disorder.

Why This Research Matters

Identifying brain-based markers that predict who is at risk for problem cannabis use could eventually help target prevention efforts and improve treatment by identifying biological mechanisms underlying vulnerability.

The Bigger Picture

This moves the field beyond simply asking "who uses cannabis" toward understanding the neural circuitry that makes some users vulnerable to problematic use. The specificity for cannabis over alcohol suggests substance-specific risk pathways in the brain.

What This Study Doesn't Tell Us

The clinical sample was small (33 participants) and predominantly male (91%). The reward task captures one aspect of brain function. Neuroimaging studies require replication in larger, more diverse samples.

Questions This Raises

  • ?Could this neural network be used as a screening tool before cannabis use begins?
  • ?Can interventions be designed to target this specific brain circuitry?
  • ?Does this network change with cannabis cessation?

Trust & Context

Key Stat:
The neural network predicted cannabis-specific risk across 3 independent samples spanning students, teens, and clinical patients
Evidence Grade:
Moderate: novel machine learning approach with external validation across multiple samples, but clinical sample was small and the network needs further replication.
Study Age:
2025 study.
Original Title:
Identification and External Validation of a Problem Cannabis Risk Network.
Published In:
Biological psychiatry, 98(8), 586-596 (2025)
Database ID:
RTHC-06948

Evidence Hierarchy

Meta-Analysis / Systematic Review
Randomized Controlled Trial
Cohort / Case-Control
Cross-Sectional / ObservationalSnapshot without intervening
This study
Case Report / Animal Study

Watches what happens naturally without intervening.

What do these levels mean? →

Frequently Asked Questions

Could a brain scan predict if someone will develop a cannabis problem?

This research identified a promising brain connectivity pattern, but it is not yet ready for clinical use as a prediction tool. More validation in larger samples is needed.

Did the same brain pattern also predict alcohol problems?

No. The identified network was specific to cannabis use outcomes and did not predict alcohol use across any of the three samples, suggesting different neural pathways for different substances.

Read More on RethinkTHC

Cite This Study

RTHC-06948·https://rethinkthc.com/research/RTHC-06948

APA

Lichenstein, Sarah D; Kiluk, Brian D; Potenza, Marc N; Garavan, Hugh; Chaarani, Bader; Banaschewski, Tobias; Bokde, Arun L W; Desrivières, Sylvane; Flor, Herta; Grigis, Antoine; Gowland, Penny; Heinz, Andreas; Brühl, Rüdiger; Martinot, Jean-Luc; Paillère Martinot, Marie-Laure; Artiges, Eric; Nees, Frauke; Orfanos, Dimitri Papadopoulos; Poustka, Luise; Hohmann, Sarah; Holz, Nathalie; Baeuchl, Christian; Smolka, Michael N; Vaidya, Nilakshi; Walter, Henrik; Whelan, Robert; Schumann, Gunter; Pearlson, Godfrey; Yip, Sarah W. (2025). Identification and External Validation of a Problem Cannabis Risk Network.. Biological psychiatry, 98(8), 586-596. https://doi.org/10.1016/j.biopsych.2025.01.022

MLA

Lichenstein, Sarah D, et al. "Identification and External Validation of a Problem Cannabis Risk Network.." Biological psychiatry, 2025. https://doi.org/10.1016/j.biopsych.2025.01.022

RethinkTHC

RethinkTHC Research Database. "Identification and External Validation of a Problem Cannabis..." RTHC-06948. Retrieved from https://rethinkthc.com/research/lichenstein-2025-identification-and-external-validation

Access the Original Study

Study data sourced from PubMed, a service of the U.S. National Library of Medicine, National Institutes of Health.

This study breakdown was produced by the RethinkTHC research team. We analyze and report published research findings without making health recommendations. All interpretations are based solely on the published abstract and study data.