Predicting relapse to cocaine-seeking behavior: a multiple regression approach

Recovery of previously extinguished responding to stimuli paired with the administration of drugs of abuse is becoming a widely utilized animal model of relapse to drug-seeking behaviors. While this approach is useful for identifying factors such as conditioned stimuli that are associated with drug-seeking, it has not directly identified behaviors that might predict susceptibility to relapse. In this study, rats were initially screened for locomotor activity in response to a novel environment. Rats were then trained to self-administer cocaine. A stimulus light and tone were paired with each infusion of cocaine. After 14 days of self-administration (maintenance) rats underwent 7 days of extinction trials (extinction phase 1), in which responding yielded neither cocaine nor the presentation of the conditioned stimulus. After extinction phase 1, rats responded for presentations of the compound stimulus in the absence of cocaine (test day 1). Rats then underwent 3 more days of extinction (extinction phase 2). After extinction phase 2, rats were once again allowed to self-administer cocaine (test day 2) and received presentations of the compound stimulus. Hierarchical regression equations, utilizing data from locomotor screening and the average responding during maintenance and extinction phases 1 and 2, were then constructed in order to predict the magnitude of responding on test days 1 and 2. A model utilizing locomotor activity, maintenance responding and extinction phase 1 responding accounted for over 76% of the variance associated with responding on test day 1, with extinction phase 1 as best predictor. A further model indicated that locomotor activity was the best predictor of propensity to self-administer cocaine on test day 2. These regression models provide a novel approach to determining factors that may predict relapse to drug-seeking behaviors.