Identification of gene (G)-environment (E) interactions is critical for understanding disease etiology, developing risk prediction models, and evaluating the impact of lifestyle interventions. Investigations of G-E interactions have led to limited findings to date, possibly due to low power for individual variants. Consequently, polygenic risk scores (PRS) are increasingly being used to detect global patterns of interaction. Motivated by the case-only method for evaluating interactions between a single variant and E, we propose a case-only method for the analysis of PRS-E interactions in case-control studies. We show that if the PRS and E are independent, a linear regression of the PRS on E in a sample of cases can be used to estimate the interaction parameter. Simulation studies indicate the efficiency of this approach is similar to that of a cohort study and about twice that of standard logistic regression analysis. Furthermore, if genotype data are available on a representative sample, the proposed method can estimate the main effect of the PRS. Extensions are considered to account for G-E dependence. We apply the proposed method to data from the UK Biobank study.