The aim of the study was to report on the clinical utility of naturalistic adjunctive treatment with valproate (VPA) in a group of panic disorder (PD) patients with comorbid bipolar disorder (BD) or otherwise resistant to antidepressants. The hypothesis was that these patients might not respond because of coexisting low-grade mood instability and adjunctive VPA treatment might ameliorate PD symptoms. A group of 47 patients with lifetime comorbid BD (n = 35, 74.5%) or otherwise resistant to antidepressants (n = 12, 25.6%), from a population of 326 consecutive outpatients with PD-Agoraphobia evaluated and treated at the Psychiatric Institute of the University of Pisa from 1991 to 1995, and followed for a period of 3 years. All patients were evaluated at baseline and at least every 2 months by means of an intensive interview including semi-structured and structured instruments (SCID, Life-Up, and Panic Disorder/Agoraphobia Interview). Mean dosage was 687 (SD = 234) mg/day (min 400, max 1,500 mg/day). Adjunctive treatment with VPA was well tolerated by all subjects, and there was no treatment interruption because of side effects or adverse events. All antidepressants-resistant subjects and 31 of 35 (88.6%) patients with bipolar comorbidity achieved symptomatological remission. During the observation period, 7 (58.3%) among resistant subjects and 17 (48.6%) of bipolar patients had a relapse of panic disorder after remission. Survival analysis of remission durations and onset relapses for PD and Agoraphobia did not show significant differences between the two groups. Relapses of Agoraphobia were less frequent and more delayed than those for panic. According to the results, VPA seems to be an effective and a well-tolerated adjunctive treatment in PD patients who were resistant to antidepressant therapy or had BD in comorbidity. The results of the study support the hypothesis of resistance to antidepressant treatment being related to mood instability.