Individualized relapse prediction: Personality measures and striatal and insular activity during reward-processing robustly predict relapse

Abstract

Background: Nearly half of individuals with substance use disorders relapse in the year after treatment. A diagnostic tool to help clinicians make decisions regarding treatment does not exist for psychiatric conditions. Identifying individuals with high risk for relapse to substance use following abstinence has profound clinical consequences. This study aimed to develop neuroimaging as a robust tool to predict relapse.

Methods: 68 methamphetamine-dependent adults (15 female) were recruited from 28-day inpatient treatment. During treatment, participants completed a functional MRI scan that examined brain activation during reward processing. Patients were followed 1 year later to assess abstinence. We examined brain activation during reward processing between relapsing and abstaining individuals and employed three random forest prediction models (clinical and personality measures, neuroimaging measures, a combined model) to generate predictions for each participant regarding their relapse likelihood.

Results: 18 individuals relapsed. There were significant group by reward-size interactions for neural activation in the left insula and right striatum for rewards. Abstaining individuals showed increased activation for large, risky relative to small, safe rewards, whereas relapsing individuals failed to show differential activation between reward types. All three random forest models yielded good test characteristics such that a positive test for relapse yielded a likelihood ratio 2.63, whereas a negative test had a likelihood ratio of 0.48.

Conclusions: These findings suggest that neuroimaging can be developed in combination with other measures as an instrument to predict relapse, advancing tools providers can use to make decisions about individualized treatment of substance use disorders.

Keywords: Methamphetamine dependence; Neuroimaging; Relapse; Reward; Risk prediction; Striatum.

Figure 1

Group by reward interaction in the striatum and insula. The linear mixed effects model revealed a significant group by reward size effect in the right striatum and left anterior insula. The group that remained abstinent showed greater activation for a large, risky relative to a small, safe rewards, while the group that relapsed showed decreased activation during large, risky relative to small, safe rewards. The right putamen cluster overlaps substantially with the brain region identified in the random forest model as predicting relapse status. Bars represent mean and error bars represent SEM.

Figure 2

Random Forest measures. These graphs show the central tendency and variance of the variables in the random forest models. Panel A shows that MD who remained abstinent showed greater activation during large, risky versus small, safe wins. MD who relapsed, in contrast, showed less differential activation when receiving large, risky versus small, safe wins. Panel B shows the values for the personality measures. Relapse and abstinent MD showed similar levels on these variables, but in combination the variables were useful in random forest modeling of relapse. This suggests that the random forest model may be able to detect higher order interactions not evident from the individual variables. Bars represent mean and error bars represent SEM.

Figure 3

Predictive value of models. In panel A, a Bayes nomogram is depicted for each random forest model. The left side of the nomogram shows the prior probability of relapse, or the proportion of the total sample that relapsed. The right side shows the posterior probability of relapse given a positive or a negative test result in the random forest model. The brackets around the central estimate represent the 95% confidence interval of the probability. When the 95% confidence intervals do not intersect, positive and negative tests are statistically significantly different. The middle line represents the likelihood ratio of a positive or negative test. All three models produced similar nomograms. In panel B, the receiver operating characteristic curves are depicted for each random forest model. All three models show significant improvement relative to the no-discrimination line.