# Neural Response to Theta‐Burst Stimulation Predicts Long‐Term Relapse in Patients With Alcohol Use Disorder: A Pilot fMRI Study

## TL;DR

This study shows that brain responses to a specific type of stimulation can predict long-term relapse in alcohol use disorder patients.

## Contribution

The study introduces a novel method using neural responses to predict relapse and evaluate treatment effectiveness in AUD patients.

## Key findings

- Active cTBS reduced relapse risk significantly compared to the sham group over 12 months.
- A machine learning model accurately predicted relapse based on brain activity changes (78.7% accuracy).
- Increased reactivity in the left medial prefrontal cortex was the strongest predictor of relapse.

## Abstract

Alcohol use disorder (AUD) is characterized by high relapse rates, and relapse is often driven by cue‐induced cravings linked to prefrontal–subcortical network dysregulation. This study investigated the neurobiological effects of inhibitory continuous theta‐burst stimulation (cTBS) targeting the right dorsolateral prefrontal cortex (rDLPFC) in patients with AUD. In a randomized, double‐blind, sham‐controlled trial, 28 patients (16 in the active cTBS group and 12 patients in the sham group) underwent 10 sessions of rDLPFC‐cTBS. fMRI was performed before and after intervention to assess neural responses to alcohol cues, and relapse was monitored for 1 year. The active cTBS group exhibited a significantly lower relapse risk over the 12‐month follow‐up compared to the sham group (HR = 0.210, 95% CI [0.070, 0.633]). A significant group‐by‐intervention interaction was found in the right superior frontal gyrus (p = 0.047); active cTBS prevented the cue‐induced hyperactivity that was observed in the sham group, suggesting a network stabilization effect. Furthermore, a machine learning model that was trained on intervention‐induced changes in brain‐wide neural activity accurately predicted long‐term relapse (accuracy: 78.7%; AUC: 0.903). Increased postintervention reactivity to cues in the left medial prefrontal cortex was the strongest predictor of relapse. These findings demonstrate that rDLPFC‐cTBS modulates craving‐related circuits and that the dynamic neural response to treatment is a powerful biomarker for predicting relapse; the findings pave the way for the development of personalized addiction medicine.

This study investigates the effects of active versus sham intervention on a specific measure (Beta value) before and after the treatment. A significant difference was found between the two groups, with a statistical result of F = 4.347 and p = 0.047. The findings indicate a measurable change from the pre‐test to the post‐test condition following the active intervention.