Firstly, the present study demonstrated that a CBT program for SAD, originally developed in Western countries, is feasible and effective in a Japanese clinical setting. To date, cultural impact is said to be relevant to the prevalence, presentation, and treatment of SAD [2
]. In Japan, for example, TKS is said to be a Japanese conceptualization of SAD. Although TKS and SAD are both characterized by excessive fear and avoidance of social interaction and performance, the concerns in TKS center on offending or embarrassing others [32
]. Despite a better understanding of the variation in symptom presentation of some of the trans-cultural variants of SAD, such as TKS, little is known about the usefulness of Western psychological therapies for it. In addition, various methodological explanations can be offered to explain cross-cultural differences or the absence of expected findings. For example, not only the use of assessment instruments but the treatment components which were developed in English and translated into Japanese may affect the outcome of CBT program. Therefore, it is necessary to investigate whether an equal treatment outcome can be achieved when transporting the Western developed CBT to Japan. Irrespective of cultural background, our results suggested that the CBT program appears to be equally acceptable to Japanese patients with SAD as to Western patients, because our dropout rate (12.3%) is generally lower than those reported in Western studies of group CBT (see Table ) and comparable to some individual CBT programs [33
Comparison of effect sizes by group CBT
In terms of effectiveness, our CBT program also compares favorably with Western reports. For example, our group CBT program was able to reduce most of the symptom measures by 20 to 30%, figures comparable to those reported in Western settings [33
]. Several previous meta-analyses of CBT for SAD have derived effect sizes based on within-group change from pre to post treatment between 0.51 to 1.06 for completers [5
]. Within-group effect sizes in our program were largely consistent with these figures (see Table ). However, effect sizes reported in the previous studies were calculated by various methods from various outcome measures. For a direct comparison with studies of group CBT conducted recently using similar SAD symptomatological scales, we calculated the effect sizes based on these scales using the formula Mpretest
. As shown in Table , our pre to post treatment effect sizes were superior to those of the previous studies of group CBT with regard to FQ-sp, SPS, SIAS, and subscales of LSAS (Fear of performance, Avoidance of performance, Avoidance of social interaction). However, two recent RCTs demonstrated superiority of individual format CBT over group format [33
]. Our program was carried out in groups of only 3 to 4 patients in order to allow as much individual attention as possible while retaining the advantage of group settings (e.g. role plays and learning through peers) and therefore can be thought to be more individualized than most reported group CBT programs (6–8 patients per group, and fixed sessions). However, the existence of a similar advantage of strictly individual format among Japanese patients has yet to be investigated.
Moderate effect sizes were demonstrated for WHLS, indicating that the impact of SAD on patients' daily lives had reduced. However, there are very few studies that used WHLS as an outcome measure, so that a direct comparison was not possible here. Similarly, a direct comparison of symptomatic remission with prior studies was also not possible because of imprecise description in many of the studies.
In addition to the outcomes described previously, we conducted a preliminary examination on whether there are any differences between patients with SAD and TKS. Six patients (10.5%) met the criteria of TKS and there were no statistical and substantial differences between patients with SAD and TKS in terms of LSAS, SPS or SIAS both before and after the treatment (For TKS, the means (SD) of LSAS, SPS and SIAS were 75.5 (43.5), 35.3 (17.4), 45.8 (19.1) at baseline, and 61.3 (49.9), 33.8 (21.8) and 43.0 (24.0) at end of treatment).
Secondly, identifying possible predictors of treatment response in the treatment completers showed that none of the pretreatment variables were significant predictors of scores of LSAS at post treatment. A number of studies have examined the role of particular variables in predicting response to treatment and their influence on overall therapeutic outcome but the results were inconsistent and inconclusive [36
]. The fact that no predicting effect emerged indicates that pretreatment variables may have little to offer for the prediction of treatment outcome. However, there are some other variables we did not directly measure in our study. For example, expectancy for improvement is said to be related to outcome [36
]. Future studies should also focus not only on pretreatment variables but also processes during the treatment, such as homework compliance and the client-therapist relationship which was suggested by Scholing et al.
The present study has several limitations. First, this is a preliminary study which was conducted in a Japanese routine clinical setting. As we modified and improved the treatment program based on the latest theoretical model, it may raise a problem of treatment consistency. A further study should examine whether there are differences in treatment outcome due to different CBT models. Moreover, lack of supervision may also have decreased treatment consistency and adherence to treatment. In Japan, there are not enough professional therapists for CBT available in routine medical settings. Regular supervision from experts in cognitive therapy is necessary and will further strengthen the training system.
Second, it must be pointed out that in the present study the principle therapist conducted the diagnostic assessments before and after the treatment, which may cause detection bias as well as performance bias. Independent blind assessor ratings would have strengthened our conclusions. Moreover, repeated t-test for comparing pre-treatment with post-treatment scores may have increased the risk of Type I error. Thus, some non-important differences may have been falsely found to be statistically significant. However, the magnitude of treatment effect was quantified by effect size as well as the percentage of reduction and the results are clearly demonstrated.
Third, the inclusion of subjects using psychotropic medication, although increasing the generalizability of the study, also confounds the treatment results because ethically we could not prohibit medication changes while subjects participated in the study. However, previous researches have failed to find a significant advantage for combined pharmacotherapy and CBT [10
]. Rosser et al.
] investigated the impact of preexisting antidepressant use on the outcome of group CBT for SAD in a naturalistic routine clinical setting and also failed to find a significant difference between combined treatments and CBT alone. In our study, none of the pretreatment variables were significant predictors of scores of LSAS at post treatment suggested that medication may have little impact on the treatment outcome. Furthermore, for those 25 (50%) patients who completed the treatment and who had already been on medication upon study participation, we examined the medication regimen at the end of treatment. Twelve (25%) of them were stabilized on the regimen throughout the treatment, while 8 (16%) had their medication reduced, and 3 (6%) had it increased. No data was available for two of the patients because they had their prescription filled elsewhere. In other words, over 90% of the subjects had their medication status unchanged or decreased through the study.
Nevertheless, considering that antidepressant medication and benzodiazepines are effective for the treatment of SAD [15
], the current study did not answer questions as to whether the treatment outcome is due to the contribution of medication or CBT alone. Future controlled trials need to explore the issue whether pre-existing medication has a significant impact on treatment outcome and also to address the question of whether treatment effects are maintained over the longer term to an equivalent extent for CBT alone, pharmacotherapy alone, or a combined approach.
Fourth, our sample size was relatively small and thus power was limited, especially with regard to predictive variables. Moreover, lack of control group data limits the generalizability of the results and lack of follow-up data limits the ability of the study to comment on longer term outcomes. Future studies including larger sample and control group should provide more insight into the CBT treatment in Japanese routine medical settings.
Despite the limitations, this pilot study provided an independent CBT treatment within a routine psychiatric service in a different cultural setting from the Western countries. It reported a preliminary treatment outcome in Japan in comparison with those from Western countries. Although there is still room for improvement, our results suggested a general replication of CBT for SAD in Japan.