OBJECTIVES: The majority (50–80%) of patients with generalized anxiety disorder (GAD) suffer from a variety of sleep problems. The objective of this study was to measure the direct effect of anxiolytic treatment and, separately, its indirect effect mediated by anxiety symptoms on sleep disturbance in patients with GAD. METHODS: Data were obtained from an 8-week, double-blind, randomized, flexible-dose, placebo- and active-control study in 372 GAD patients. Patients were assigned to pregabalin (300–600 mg/day), venlafaxine XR (75–225 mg/day), or placebo.  Anxiety symptoms were measured by clinician-administered HAM-A scale at baseline and on weeks 1–4, 6, and 8 and sleep disturbances were measured using the 4-item sleep disturbance subscale on Medical Outcomes Study Sleep Scale (MOS-SS), a validated patient-reported outcome measure. Statistical mediation modeling was used to estimate the direct effects of pregabalin and venlafaxine XR (relative to placebo) and the indirect effects via anxiety symptoms, the mediator variable, as measured by HAM-A total score. All available data from the trial were used in the statistical analyses. RESULTS: Patients were predominantly female (61%) and had a mean age of 41 years. Path coefficients for direct and indirect (mediated) paths for patients in the pregabalin group indicated less sleep disturbance with treatment. Fifty-three percent of the reduction in sleep disturbance was due to the direct effect of pregabalin (p=0.007). The remainder of the reduction in sleep disturbance (47%, p=0.015) was mediated via anxiety symptoms. In contrast, there was no difference in sleep disturbance (p=0.58) between patients treated with venlafaxine XR or placebo. CONCLUSIONS: The underlying relationships between treatment, anxiety symptoms, and sleep disturbance are explicitly described and partitioned through mediation modeling.  About half of the total improvement in sleep disturbance was identified as being due to the direct effect of pregabalin independent of an effect on anxiety symptoms.