We investigated dimensions of liability to Generalized Anxiety Disorder (GAD) and whether evidence exists for distinct pathological versus normal clusters in the population. Structured interviews were administered to a general population sample of 2,163 female twins in a cross-sectional design. Endorsement rates were estimated using full information maximum likelihood factor analyses of the DSM-III-R and DSM-IV GAD symptoms, which provides appropriate treatment of the stem-probe structure of the clinical interview. Endorsement rates were highest for symptoms retained in DSM-IV. For both DSM-III-R and DSM-IV, a two-factor model fit the data better than a single-factor model. There was no evidence for non-normality in the liability to GAD. For DSM-III-R, autonomic symptoms loaded on a factor with panic disorder, while fatiguability, difficulty concentrating and hypervigilance loaded on a factor with major depression. For DSM-IV, all items loaded on one factor, and muscle tension also loaded on a second. Major depression, panic, phobias and alcohol dependence diagnoses also loaded on the first factor.
Conclusions: Future research involving structured interviews should take into account the stem-and-probe format and focus on common factors rather than separate disorders; GAD is not a unidimensional construct and pathological anxiety may differ only quantitatively from normal anxiety.