Fifty seven consecutive patients with SAD were recruited into the outpatient group-based CBT program at the Department of Psychiatry, Nagoya City University Hospital, Japan between July 2003 and January 2007. Some patients were referred from mental health professionals and others sought treatment for social anxiety disorder on their own.
All patients were diagnosed with SAD as the primary disorder according to the DSM-IV criteria, as assessed by the Structured Clinical Interview for DSM-IV . In addition, all patients fulfilled the following criteria: (a) absence of a history of psychosis or bipolar disorder, and of current substance use disorder; and (b) no previous CBT treatments, nor any other additional psychosocial therapies during the treatment, and (c) absence of Cluster B personality disorder. With the exception of criterion (c), current Axis I major depressive disorder, other current anxiety disorders and Axis II personality disorders were not a reason for exclusion if the patients' symptoms abated enough to allow them to attend the group CBT sessions regularly.
The patients provided their written informed consent after full explanation of the objectives and procedures of the present study. The study's protocol was approved by the Ethics Committee of Nagoya City University Graduate School of Medical Sciences.
Initially we started the program with 12 two-hour sessions, adopting the model developed by Gavin Andrews et al. in Sydney, Australia. As we modified and improved the program according to Clark & Wells' model , the program was extended to 12 or more sessions depending on each group's needs. After this modification, the average number of sessions per group was 15 (range: 11 to 20).
The program is based on Clark & Wells' model, which focuses on the maintenance of SAD. The main components include 1) psychoeducation regarding the nature of anxiety and symptoms of SAD, 2) deriving an individualized version of the cognitive-behavioral model of SAD using patient's own thoughts, images, anxiety symptoms, safety behaviors and strategies of attention, 3) experiment to drop safety behaviors and self-focused attention, 4) attention training to shift focus of attention externally to the task at hand or the social situation, 5) video feedback of role-plays of moderately anxiety-provoking situations to modify distorted self-imagery, 6) behavioral experiments to test the patients' catastrophic predictions during planned exposure to the situations in within-session role-plays and in vivo homework assignments, and 7) cognitive restructuring for identifying and modifying dysfunctional assumptions especially during behavioral experiments. Homework was assigned after every session.
The patients were treated in groups of 3 or 4 led by two therapists (one principal therapist and one co-therapist). This format was adopted because we found it was too difficult to provide each one out of four or more patients with the individual attention that is needed for the detailed practices of CBT techniques. Most of the patients appeared to be comfortable with this format. Eight therapists (five psychiatrists and three doctoral-level clinical psychologists) with more than three years of clinical practice and experience in treatment of anxiety disorders conducted the treatment program guided by a therapist manual in Japanese. During the treatment, therapists had group discussion once a month to check adherence to the program and to plan for future sessions. However, the quality of treatment delivery was not rated though audio or video recordings.
During the CBT treatment, co-administration of antidepressants and benzodiazepines was allowed because there is no evidence to contraindicate their concurrent use [15, 16], and also because symptoms of other comorbid anxiety and/or depressive disorders had to be controlled with antidepressants.
Patients were assessed with a structured diagnostic interview as well as an extensive questionnaire battery. At baseline, before treatment commencement, the principal therapist in charge of the CBT group administered the mood and anxiety disorder sections of the Structured Clinical Interview for DSM-IV , in order to ascertain SAD diagnosis and any mood and anxiety comorbidities. In addition, eight self-reported and one clinician-administered questionnaires to measure various aspects of SAD symptomatology and functional impairment, broader psychopathology in general, and premorbid personality were administered at the same time. The same instruments, except for the personality inventory, were repeated at the end of the program.
Social Phobia Scale and Social Interaction Anxiety (SPS/SIAS)
The SPS/SIAS  is a 20-item self-report questionnaire respectively. The SPS was designed to measure the fear of being observed whereas the SIAS provides a measure of fear of social interaction. Each scale has twenty items rated on a 4-point scale, from 0 (not at all characteristic or true of me) to 4 (extremely characteristic or true of me), with scores for each scale ranging from 0 to 80. Excellent internal consistency reliability and sufficient discriminant, predictive and concurrent validity have been demonstrated for both of the Japanese versions .
Fear Questionnaire – social phobia subscale (FQ-sp)
The FQ-sp is a 5-item self-reported instrument for measuring the fear-motivated avoidance of being observed, performing, being criticized, and talking to authorities . Items are rated on a 9-point Liker-type scale, from 0 (would not avoid it) to 8 (always avoid it). Good test-retest reliability and factor validity have been demonstrated [19, 20].
The Fear of Negative Evaluation Scale (FNE) and Brief Fear of Negative Evaluation Scale (BFNE)
The FNE is a 30-item questionnaire developed to measure fear of negative evaluation in social situations . Items are rated as true or false, with the total score ranging from 0 to 30. Good reliability and validity of the Japanese version have been reported . However, the newer brief version of the FNE contains 12 items with a 5-point Likert-type format (1 = "not at all characteristic of me"; 5 = "extremely characteristic of me") has been found to be more sensitive to treatment-related change . We therefore switched to BFNE after several groups. As the BFNE was not given to patients from the beginning, calculations can only be made for a smaller sample of n = 37 (FNE) and n = 26 (BFNE).
The Liebowitz Social Anxiety Scale (LSAS)
The LSAS is the most frequently used clinician-administered instrument for the assessment of social phobia. It is a 24-item scale that provides separate scores for fear (0–3 indicate none, mild, moderate, and severe, respectively) and avoidance (0–3 indicate never, occasionally, often, and usually, respectively) of social interaction and performance situations, thus providing scores of four subscales, Fear of Performance, Avoidance of Performance, Fear of Social Interaction, and Avoidance of Social Interaction. The Cronbach's alpha reliability estimate of the Japanese version is 0.95 for total items and 0.80–0.91 for subscales. In addition, sufficient validity data have also been provided .
Symptom Checklist-90-Revised (SCL-90-R)
The SCL-90-R is a 90-item questionnaire widely used for the assessment of general psychopathology . It yields scores for nine primary symptom dimension of somatization, obsessive-compulsive, interpersonal sensitivity, depression, anxiety, hostility, phobic anxiety, paranoid ideation, and psychosis. The reliability and validity of the Japanese version have been reported .
Work, Home Management, Social and Private Leisure Activities Scale (WHLS)
The WHLS is a self-report scale for measuring functional impairments in the areas of work, home management, social and private activities . The patients' answers are graded between 0 (not at all impaired) to 8 (very severely impaired). Satisfactory reliability and construct validity have been reported .
NEO Five Factor Index (NEO-FFI)
This is a 60-item self-reported questionnaire designed to measure the five major personality dimensions of neuroticism, extraversion, conscientiousness, openness and agreeableness . The Japanese version has been tested in a general population and satisfactory internal consistency (coefficient alpha) is provided as well as validity .
First, treatment completers were compared with patients dropping out during treatment using unpaired t tests for continuous variables or chi-squared tests for categorical variables.
Second, the outcomes of the CBT program for the Japanese patients with SAD were quantified in the following ways. Our primary outcome was the total LSAS score, and percentages of patients with symptomatic response and remission were calculated. Following Bandelow et al., treatment responder status is defined as 31% or greater reduction in the LSAS and remission as a score of 36 or less on the LSAS.
Furthermore, to examine the outcomes of the CBT program across various aspects of the disorder, pre- and post-treatment scores were compared for each SAD symptomatological scale (SPS, SIAS, FQ-sp, FNE, BFNE, LSAS) as well as the functional scale (WHLS) using paired t tests. The magnitude of treatment effect was quantified in two ways by way of the percentage of reduction ((pre – post-treatment)/pretreatment scores × 100) as well as the effect size ((M
). According to Cohen , effect sizes are categorized as follows: small (0.20–0.49), medium (0.50–0.79), and large effects (0.80 and above).
All statistical analyses for these treatment outcomes were conducted twice, once based on the intention-to-treat (ITT) principle whereby all the dropouts were considered to be non-responders and their last available observations were carried forward, and once among the completers only. Treatment completers were defined as participants who had attended at least 80 percent of the treatment sessions and returned post treatment questionnaires. The ITT analyses present the more conservative estimates of treatment effectiveness.
Third, in order to elucidate the baseline predictors of the treatment outcomes, multiple regression analyses were conducted with the post-treatment LSAS score as a dependent variable and the baseline demographic and clinical variables as independent variables, while controlling for the baseline LSAS scores. These analyses were performed among the completers only.
All the statistical tests were two-tailed, and an alpha value of less than 0.05 was considered statistically significant. All the data were examined using SPSS 15.0 for Windows [SPSS Inc., 2007].