Further development of a neural network for panic disorder

Previously, we had developed a neural network model of panic disorder based on the neuroanatomic hypothesis advanced by Gorman et al (1989). The model is able to simulate both normal and paroxysmally severe levels of anxiety. Experimental manipulations of the network were also able to simulate the effects of serotonergic medications and cognitive-behavioral therapy as treatments in reducing the frequency of panic attacks. Gorman et al, based on their neuroanatomic analysis, made predictions about the nature of panic attacks and their response to treatment. This study utilized an expanded version of the neural network model already developed to test several of Gorman et alʹs predictions. Results partially confirmed the predictions that were amenable to evaluation with this model. Anxiolytic treatments did not completely block panic. Benzodiazepine treatment reduced anticipatory anxiety but not panic. Reduction in agoraphobic symptoms did not completely eliminate panic as well. These results confirm the potential usefulness of neural network modeling in evaluating hypotheses about neurophysiologic function in psychiatric disorders. Results provide additional support for the predictions made by Gorman et al about panic disorder.