Background: Theories regarding how spontaneous panic and agoraphobia relate are based mostly on cross-sectional and/or clinic data.

Aims: To determine how spontaneous panic and agoraphobia relate longitudinally, and to estimate the incidence rate of and other possible risk factors for first-onset agoraphobia, using a general population cohort.

Method: A sample of 1920 adults in east Baltimore were assessed in 1981-1982 and the mid-1990s with the Diagnostic Interview Schedule (DIS). Psychiatrist diagnoses were made in a subset of the sample at follow-up (n = 816).

Results: Forty-one new cases of DIS/DSM-III-R agoraphobia were identified (about 2 per 1000 person-years at risk). As expected, baseline DIS/DSM-III panic disorder predicted first incidence of agoraphobia (OR = 12, 95% CI 3.2-45), as did younger age, female gender and other age, female gender and other phobias. Importantly, baseline agoraphobia without spontaneous panic attacks also predicted first incidence of panic disorder (OR=3.9, 95% CI1.8-8.4). Longitudinal relationships between panic disorder and psychiatrist-confirmed agoraphobia were strong (panic before agoraphobia OR=20, 95% CI 2.3-180; agoraphobia before panic OR=16, 95% CI 3.2-78).

Conclusions: The implied one-way causal relationship between spontaneous panic attacks and agoraphobia in DSM-IV appears incorrect.