Background. A recent meta-analysis provides evidence that generalized anxiety disorder (GAD) is familial. However, two of the key studies relied on subjects who were self-selected or recruited from the clinic setting, thereby limiting generalizability.

Method. We conducted a family study of GAD in which probands and controls came from a community sample originally enrolled in a prevalence study in Edmonton, Canada. One hundred and sixty probands, 764 controls and 2386 first-degree relatives (FDRs) were interviewed using the Diagnostic Interview Schedule (DIS); lifetime diagnoses were made according to DSM-III criteria without exclusions. Logistic regression analysis was performed with GAD (in a proband) as the ‘exposure’, and GAD in an FDR as the ‘outcome’. Several analytic strategies were used to control for potential confounding by major depressive disorder (MDD) and several anxiety disorders (panic disorder, phobic disorders, obsessive–compulsive disorder, and post-traumatic stress disorder).

Results. The odds ratios for the association between GAD in a proband and GAD in an FDR were in the range 1·4–1·8 when the entire FDR sample was analysed, and in the range 2·1–2·8 when we restricted to FDRs who were children of probands and controls.

Conclusion. In the community setting, GAD exhibits mild to moderate familial aggregation.