Despite extensive literature searches and personal communication, only two rigorous studies were identified that examined combined benzodiazepines and psychotherapy for panic disorder.
What little evidence there is suggests, however, that behaviour therapy (exposure) plus benzodiazepines may be superior to benzodiazepines alone, especially at the end of acute treatment (RR = 3.39, 1.03 to 11.21), and possibly also during the acute phase and at long-term follow-up. The combined therapy may also be superior to behaviour therapy alone during the acute phase, but this possible superiority may not persist and perhaps invert during the follow-up 6–12 months after treatment termination.
With respect to combination therapy versus benzodiazepines alone, secondary endpoints representing continuous measures also hint at the superiority of the combination during the acute phase treatment, in that the combination was significantly superior to benzodiazepines in terms of the global severity, phobic avoidance, and social functioning. Over the long term, this superiority was not statistically significant (RR = 2.31, 0.79 to 6.74). However, the secondary outcome demonstrated combination therapy was significantly superior to benzodiazepine alone. In addition, data from one of the included RCTs indicated that even placebo was not inferior to behaviour therapy alone or benzodiazepine alone for panic attacks as opposed to phobias and disability, for which placebo was indeed inferior. [20
With regard to behaviour therapy alone, some secondary endpoints hinted at the superiority of the combination during the acute phase treatment (frequency of panic attack, social functioning). Some authors have argued that the combination therapy is superior to psychotherapy alone in the long term [28
]. However the results in the present review failed to reach statistical significance in terms of response rate (RR = 0.62, 0.36 to 1.07, P = 0.08) or the secondary outcome representing global severity of panic attacks (SMD = -0.19, -0.59 to 0.22, P = 0.37).
Considering these results, benzodiazepine alone is not to be recommended for those who have access to appropriate resources of behaviour therapy.
We did not focus on the combined antidepressants and psychotherapy in the present review, but our companion systematic review on this topic also concluded that either combined therapy or psychotherapy alone rather than antidepressants alone might be chosen as first-line treatment for panic disorder with or without agoraphobia [6
]. Combining the conclusions of the two reviews, pharmacotherapy alone, be it by an antidepressant or by a benzodiazepine, may be avoided in treatment for panic disorder where psychotherapy is available.
One has to note some possible weaknesses in the present review. First of all, only two studies including 243 patients diagnosed as panic disorder with agoraphobia were identified, type II errors were therefore likely and even the significant results might have limited generalizability. Moreover, we were not able to do any analysis on panic disorder without agoraphobia and managed only to establish very limited evidence for panic disorder with agoraphobia. Secondly, generalizability of the present findings beyond specialist psychiatric settings is not straightforward. The included studies were conducted at psychiatric hospitals under psychiatrists' and clinical psychologists' administration [20
], so that generalizability of the results of the present review to primary care may be questioned. Another concern on generalizability of the review may arise from the fact that participants were limited to those having been able to tolerate being off medication before entering the study in one study [20
]. In addition, regarding use of benzodiazepines, one included study employed a huge daily dose of alprazolam in both the benzodiazepine arms [20
] compared to the dose usually recommended in clinical practice, and this could have had negative effects on these arms. The other included study [41
] used diazepam as benzodiazepine in relatively lower dose compared to previous studies which involved diazepam as treatment for panic disorder [55
]. Although diazepam has been demonstrated not only superior to pill-placebo but also equally effective compared with alprazolam [56
], this issue could have diminished efficacy of diazepam and have diluted differences between combined therapy and psychotherapy alone. Thirdly, the definition of "response" employed in the present study differs from study to study, due to reported outcomes in the original reports. Clinical global rating was employed in one study [20
] and phobic avoidance scale in the other [41
]. The proportions of responders are much smaller in one study [41
] than the other [20
], possibly due to different definitions of response. Although no significant heterogeneity has been observed through the analyses in the present review, one might think these definitions are arbitrary or plausible. However, we failed to obtain clinical global scale in the latter [41
] in spite of contacting the original authors. We therefore utilized only phobic avoidance scale to define the response, since this has been shown to better represent the patients' global status [34
]. Last but not least, all the included studies dealt with psychotherapy in the form of exposure therapy. Other forms of psychotherapies including cognitive restructuring and psychodynamic therapy also need to be examined in combination with benzodiazepines.
On the other hand, the present review may have several strengths. First, we performed a systematic and comprehensive search for the relevant trials. All references to possible candidate studies were investigated to identify relevant ones. The authors of the included studies were contacted to give more information about their studies and other possibly relevant ones. Consequently, our analyses managed to include some unpublished data (MAL) in one study [41
]. Second, though there are many narrative reviews on efficacy of combination therapy against either monotherapy that included both high-quality and low-quality studies all together, we limited our systematic review only to high quality RCTs. These RCTs were also conducted while keeping assessors blind. Our approach minimized bias and therefore offered more rigorous evidence. Third, we applied strict ITT principle, by means of imputing response rates from continuous measures of global severity for panic disorder and counting all the dropouts as non-responders, as we were interested in the fate of all patients who began the acute phase after randomisation, not just those who completed it. Counting dropouts as non-responders might appear too stringent as some dropouts may have improved, but we cannot confirm their percentage. Given this circumstance, this approach is the most conservative estimate of the response rate, and in the absence of any stronger and more rational alternative, we decided to remain conservative with regard to either treatment arm.