Bayesian computational markers of relapse in methamphetamine dependence.

Area of Science:

• Neuroscience
• Computational Psychiatry
• Addiction Medicine

Background:

• Methamphetamine use disorder (MUD) has a high relapse rate, necessitating effective relapse prediction.
• Understanding the cognitive mechanisms of relapse is crucial for developing targeted secondary prevention strategies.
• Computational approaches offer mechanistic insights into psychiatric vulnerability and predictive markers.

Purpose of the Study:

• To identify neural predictors of relapse in individuals with MUD using computational modeling.
• To investigate the role of Bayesian prediction error signaling in relapse vulnerability.
• To explore the utility of functional magnetic resonance imaging (fMRI) and the Stop Signal Task (SST) in predicting relapse.

Main Methods:

• Sixty-two individuals with MUD from an inpatient program underwent fMRI during an SST.
• Participants were prospectively followed for one year to assess relapse.
• Neural activity related to Bayesian prediction errors was analyzed and compared between abstinent and relapsed groups.

Main Results:

• Thirty-three percent of participants relapsed within one year.
• Individuals who relapsed showed significantly smaller neural activations to Bayesian prediction errors in the left temporoparietal junction, left inferior frontal gyrus, and left anterior insula compared to abstinent individuals.
• No differences in neural activation to non-model-based tasks or self-report measures were found between groups.

Conclusions:

• Bayesian cognitive models and associated neural activity can serve as predictive biomarkers for relapse in MUD.
• Deficits in belief processing and updating, as indicated by altered Bayesian prediction error signaling, may underlie relapse vulnerability in MUD.
• These findings support a computational framework for understanding and potentially preventing relapse in MUD.