Several studies have examined the predictors of treatment response in obsessive–compulsive disorder (OCD). Only limited information is available on the predictors of response to antipsychotic augmentation of serotonin reuptake inhibitors (SRIs). Data from placebo-controlled studies of augmentation with quetiapine were combined in a best subsets logistic regression to derive a predictive model for Yale-Brown Obsessive–Compulsive Scale (YBOCS) change and the YBOCS endpoint. Data from the YBOCS checklist and a variety of clinical and demographic variables previously shown to predict treatment outcome in OCD were analysed. In univariate analyses, the failure of fewer previous SRI trials was associated with the YBOCS response. In the multivariate model, for YBOCS change, 45% of the variance was attributed to the fact that patients had failed fewer previous SRI treatments, had higher baseline obsession scores, and ordering and arranging compulsions. For the YBOCS endpoint scores, 50% of the variance was attributed to the fact that patients had fewer failed SRI trials, higher baseline compulsion scores, and counting/ordering and arranging compulsions. These data indicate a number of predictors of response to augmentation of SRIs in treatment-refractory OCD. These include fewer previously failed SRI trials and generally higher overall baseline scores for obsessions and compulsions as well as counting/ordering and arranging compulsions. Other factors are, however, also likely to play an important role in predicting outcome.