The issue of panic disorder resistant to treatment (whether pharmacological or psychological) has attracted little research attention, despite its clinical frequency and importance. The aim of this study was to compare three treatment modalities, namely exposure alone (E), exposure associated with imipramine (EI) and cognitive therapy supplementing exposure (EC), in a sample of 21 patients with DSM-IV panic disorder and agoraphobia, who failed to respond to a first standard course of individual behavioural treatment based on exposure in vivo. Treatments were administered according to a cross-over, controlled design (E-EI-EC, EI-EC-E, EC-E-EI). Twelve of the 21 patients achieved remission (panic-free status) during the trial. In 8 cases this occurred after exposure alone (E) and in two cases each after the other treatments (EI and EC). The results revealed a significant effect of the factor time on a number of variables, and the superiority of exposure alone compared to other treatment modalities with regard to some variables. These findings suggest that long-term behavioural treatment based on exposure may be necessary in some patients, and may induce clinical remission. However, patients who do not respond to exposure show poor tolerance of and compliance with pharmacological treatment, and are unlikely to achieve remission with imipramine or cognitive therapy, even though this may occur in individual cases.