Category: Translational

Symposium

Bayesian Neural Adjustment of "the Need to Stop" Predicts Relapse in Methamphetamine-Dependent Individuals

Friday, November 17
12:00 PM - 1:30 PM
Location: Sapphire Ballroom M & N, Level 4, Sapphire Level

Keywords: Substance Abuse | Neuroscience | Information Processing
Presentation Type: Symposium

Background: Determining the precise nature of the cognitive dysfunctions in substance dependent individuals that contribute to relapse is essential for process-specific relapse prevention efforts.  Recently, we have implemented a computational approach based on a Bayesian ‘ideal observer’ model to better delineate the cognitive dysfunctions associated with inhibitory function.  This study investigates whether neural activation underlying those computational processes can predict the likelihood of relapse in recently sober methamphetamine-dependent individuals (MDI). 


Methods: Fifty-eight MDI were recruited from a 28-day inpatient treatment program and completed a stop-signal task while undergoing functional magnetic resonance imaging (fMRI) after approximately 3-4 weeks of sobriety.  These individuals were prospectively followed for 1 year and assessed for relapse to substance and alcohol use. A Bayesian ideal observer model was used to infer individuals’ trial-wise expectations about encountering a stop signal in the task, i.e., the “need to stop."

Results: Of the 58 MDI, 19 (33%) reported relapse one year after treatment. Neural activations associated with two types of Bayesian prediction error, i.e. the difference between actual and expected need to stop on a given trial, in the left inferior frontal gyrus (odds ratio, .27 [95% CI=0.11-0.66]; p<.01) and in the left angular gyrus (odds ratio, 7.76 [95% CI=2.35-19.04]; p<.01) were associated with a higher likelihood of relapse. Combined, these two Bayesian predictors provided a predictive accuracy of 80% (72% sensitivity and 82% specificity). In contrast, neither non-Bayesian neural variables nor self-report clinical measures were found to robustly predict relapse.

Conclusions: These results suggest that MDI exhibiting neural inefficiencies to support computational processes underlying inhibitory response prediction may be more likely to relapse. Bayesian cognitive models not only provide a computational explanation of belief processing deficits in MDI, but they may be particularly useful in predicting long-term clinical outcomes, such as relapse, in this population.

Katia M. Harlé

Assistant Clinical Professor
University of California, San Diego

Presentation(s):

Send Email for Katia Harlé


Assets

Bayesian Neural Adjustment of "the Need to Stop" Predicts Relapse in Methamphetamine-Dependent Individuals



Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Bayesian Neural Adjustment of "the Need to Stop" Predicts Relapse in Methamphetamine-Dependent Individuals