Background: Effect sizes of pharmacotherapy in alcoholism are modest. They might improve if subjects could be divided into more homogeneous subgroups and would then be treated targeted to their neurobiological profile. In such an effort, we tested neural cue reactivity as a potential predictor of treatment response to naltrexone. Alcohol-associated cues cause brain activations in mesocorticolimbic networks due to the positive reinforcing properties of alcohol. These activations were reported to be associated with relapse behavior. Naltrexone, an antagonist at the mu-opioid receptor, improves drinking behavior in some but not all patients probably by blocking the positive reinforcement of alcohol. Conversely, acamprosate is proposed to modulate negative reinforcement (withdrawal and cue-induced withdrawal). Identifying subjects with elevated cue reactivity and testing their response to medical treatment could thus improve our understanding of some of the mechanisms underlying pharmacotherapy response. Methods: A picture-perception task featuring alcohol-related and neutral stimuli was presented to 64 recently detoxified alcohol-dependent patients. Patients came from 1 center of a larger double-blind randomized multicenter clinical trial (the "PREDICT Study"). They were scanned prior to being randomized to either naltrexone or acamprosate. We examined the interaction between medication and functional magnetic resonance imaging (fMRI) cue reactivity, as measured by the percentage of voxels activated, using the time to the first severe relapse as the outcome criterion. Our a priori formulated hypothesis was that naltrexone but not acamprosate should be efficacious in subjects with high cue reactivity. Results: We observed an interaction effect between pretreatment brain activation induced by alcohol images and medication (acamprosate/naltrexone) on relapse behavior. In line with our hypothesis, this interaction was driven by treatment response to naltrexone in patients with elevated pretreatment cue reactivity in the ventral striatum. Conclusions: fMRI has the potential for predicting treatment response to naltrexone in a subgroup of alcohol-dependent patients. However, this approach will be limited to researching the mechanisms and principles of treatment response.