Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Psychotherapy and antidepressant for panic disorder

Systematic review

Panic disorder, with or without agoraphobia, is usually treated with psychotherapy, with or without antidepressants. Deciding which is best is made difficult by trials being generally small, testing different types of psychotherapy, with different controls, and with endpoints at the end of treatment or at some later time. Just as well that we have a systematic review [1] to help us get some grip.

Systematic review

Searching used a specialised Cochrane register, supplemented with a series of other electronic searches, plus hand searching of major journals, proceedings, and trial registers, up to the end of 2003. Randomised trials were accepted. The primary outcome was response, a substantial improvement from baseline, equivalent to very much or much improved, derived by dichotomising results from studies which used different measurement scales. Outcomes were assessed at the end of the treatment phase, and at some longer follow up, typically 12-24 months later.


Searching found 23 randomised comparisons with 1,700 patients, in studies with treatment duration of six to 26 weeks, but where most were of nine to 15 weeks. Different types of psychotherapy were used. Behaviour therapy consisted of exposure, with or without breathing retraining, with or without relaxation exercises. Cognitive behavioural studies consisted of both cognitive and behavioural elements. Antidepressants used included tricyclics (average dose 150 mg daily of imipramine equivalents), or SSRIs (average dose 30 mg daily of fluoxetine equivalents).

The summary results are shown in Table 1. Overall, psychotherapy plus antidepressant was better than antidepressant or psychotherapy alone at the end of the treatment phase, with numbers needed to treat of 10 to produce one more responder. At longer term follow up, psychotherapy plus antidepressant was better than antidepressant alone, with an NNT of 5, but not psychotherapy alone. Figure 1 shows the percentage of responders in the comparisons in a truncated form, to emphasis the similarity between psychotherapy plus antidepressant.

Table 1: Summary of pooled estimates of efficacy of psychotherapy plus antidepressants versus antidepressants alone or psychotherapy alone in treating panic disorder

Number of
Responders (%) with
Relative risk
(95% CI)
(95% CI)
End of treatment phase
Psychotherapy + antidepressant vs antidepressant
1.2 (1.1 to 1.4)
10 (6 to 37)
Psychotherapy + antidepressant vs psychotherapy
1.3 (1.1 to 1.4)
10 (6 to 23)
Longer term follow up
Psychotherapy + antidepressant vs antidepressant
1.6 (1.2 to 2.1)
5.6 (3.7 to 12)
Psychotherapy + antidepressant vs psychotherapy
0.9 (0.8 to 1.2)
not calculated

Figure 1: Percentage of responders in comparisons of psychotherapy plus antidepressants (PA) versus antidepressants alone (A) or psychotherapy alone (P)

Various sensitivity analyses were performed. Omitting studies that included agoraphobia made no difference, nor did comparison with different classes of antidepressant, nor did analysis by different type of psychotherapy make much difference. Additional analysis by size of trial made no difference. Sensitivity analyses like this are important, given the small size of trials generally, and the resultant scattering of results, possibly because of random chance. Figures 2 and 3 show results for individual trials comparing psychotherapy plus antidepressant with antidepressant at the end of treatment (Figure 2) and at later follow up (Figure 3).

Figure 2: Responders at end of treatment phase with psychotherapy plus antidepressants (PA) or antidepressants alone (A)

Figure 3: Responders at longer term follow up with psychotherapy plus antidepressants (PA) or antidepressants alone (A)


These results appeared pretty robust. One possible quibble is that the numbers given in the paper for psychotherapy plus antidepressant compared with antidepressant alone at the end of treatment had some obvious minor problems, like numbers not adding up, and relative risks which were obviously wrong, but recalculating or omitting problematic trials made no difference. The results were robust.

Bandolier is always suspicious about results when we have only a scattering of small studies, and where there is scope for clinical heterogeneity. That was the situation here, and as we saw in Bandolier 139, that is a situation where even a systematic review is more likely to be wrong than right. That is why sensitivity analysis is important, and Figures 2 and 3 emphasise how much small trials can contribute to an overall result.

Here the results for psychotherapy plus antidepressant stood up to sensitivity analysis. The question of whether the extent of advantage over antidepressant alone is big enough, or big enough to make a cost effectiveness argument, is another matter, and a difficult call. Two years after the event some 60% of patients with panic disorder had not achieved a sustained response with any of these treatments.


  1. TA Furukawa et al. Psychotherapy plus antidepressants for panic disorder with or without agoraphobia. British Journal of Psychiatry 2006 188: 305-312.

previous or next story