Abstract
Social phobia is characterised by a marked and persistent fear of one or more social or performance situations and the fear that the patient will act in a way that will be humiliating or embarrassing [1]. Patients with social phobia attempt to avoid the feared situations when they can. Very often, however, they have little choice but to endure anxiety provoking situations, particularly if the negative consequences of non-attendance outweigh the negative consequences of attendance [2].
Patients with social phobia are exposed to social situations daily but often do not experience reduction in anxiety [3–4]. Such a reduction in anxiety would be predicted by learning theory models. The absence of significant reductions in anxiety, despite repeated exposure, may in part be due to the individual's utilisation of ‘safety behaviours’ which defer or deflect actual exposure to the feared situations [4–9].
The term safety behaviour describes ‘…escape behaviours and subtle avoidance, but also…general avoidance’ [9, p.33]. Salkovskis [9] hypothesised that engaging in safety behaviour subjectively ‘saves’ the person from perceived threat. For the patient, the safety behaviour has a logical link to the perceived threat. Being saved strengthens the person's belief that their safety behaviour prevented likely danger but makes disconfirmation of threat-related cognitions unlikely. By explaining this logical link based on a cognitive model, patients understand why it may be helpful to stop engaging in safety behaviours.
Controlled studies in the literature give conflicting views as to whether adding cognitive strategies to exposure-based therapies improves treatment outcome. Butler [3] noted that exposure alone for social phobia did not significantly alter the associated underlying fear of negative evaluation. Additionally, an earlier study [10] indicated that, although patients who were treated with exposure alone improved on measures of phobic severity and social avoidance, 40% of these patients requested further treatment the following year. Butler et al. [10] concluded that, without addressing the underlying fear of negative evaluation, patients were likely to continue to seek help. Emmelkamp, Mersch, Vissia and van der Helm [11] showed that exposure alone did not alter negative cognitions in social phobia but that cognitive treatment alone did so. Similarly, Mattick and Peters [12] and Mattick et al. [13] reported that exposure alone decreased social phobic avoidance but did not result in significant changes in fear of negative evaluation. They found that combined exposure and cognitive restructuring was superior to exposure alone.
Salkovskis [9] suggests that exposure is optimised when combined with cognitive strategies. When combined, prolonged exposure would not be necessary; rather, the creation of behavioural experiments (exposure exercises) can be used to gather evidence to invalidate the feared catastrophe [5].
Clark and Wells' [7] cognitive model for social phobia provides the theoretical basis for the present study. They hypothesised that the perceived danger of the social situation triggers an anxiety response with physiological, behavioural, emotional and cognitive features. The physiological anxiety symptoms themselves further exacerbate the perceived threat of danger. In their model, Clark and Wells [7] hypothesised a number of processes that maintain social phobia by preventing disconfirmation of negative cognitions about danger in social situations. First, in an attempt to prevent the feared social danger from occurring, the person engages in safety behaviours such as wearing make-up to hide blushing, or gripping cups or glasses tightly to control shaking. Mental manoeuvres, such as rehearsing sentences before speaking during conversation in order to ‘come across well’ are also common safety behaviours used by the socially phobic individual. The individual perceives that such behaviours may ‘save’ him or her from negative social evaluation. As a result, the individual does not have the opportunity to learn that the feared outcome may not occur, or that if it does occur, will not necessarily result in a negative outcome. That is, the individual ‘sabotages’ the exposure and hence does not have a chance to disconfirm the feared social danger [9].
Secondly, Clark and Wells [7] propose that when the person perceives they are under threat they also divert their attention to intensive self-monitoring and create a picture in their minds of how they believe others see them. This self-focusing heightens negative self-evaluation and prevents scanning the environment and disconfirming the interoceptive information. Finally, performance can be inhibited by anxiety and this may actually impact on the way others think and behave towards the socially anxious person.
The aim of the present study was to compare the relative efficacy of a standardised group cognitive-behavioural therapy (CBT) program [2] with a standardised group CBT program plus instruction to drop safety behaviours. It was hypothesised that instruction to drop safety behaviours would enhance the efficacy of CBT treatment.
Method
Subject selection procedures and criteria
Thirty-six patients (consecutive attenders to a tertiary referral anxiety disorders clinic) were interviewed by an experienced clinician. Those who met criteria for a primary DSM-IV[1] diagnosis of social phobia were offered treatment and randomly assigned to standard CBT group ‘USUAL’ or standard CBT group treatment plus instruction to drop safety behaviour ‘SAFETY’. Avoidant personality disorder was not an exclusion criterion. Comorbid lifetime diagnoses were generated by the Composite International Diagnostic Interview (CIDI-Auto) [14] for 28 of the 30 subjects who completed the study (one of these 28 subjects did not meet criteria for social phobia on this measure). Of those 28 subjects who did complete the CIDI-Auto, the majority of subjects met criteria for additional diagnoses in conjunction with a primary diagnosis of social phobia as shown in Table 1.
Comorbid lifetime diagnoses generated by CIDI-Auto
The 36 subjects were divided among two treatments (USUAL or SAFETY) with two groups per treatment program. Two USUAL groups initially comprised 16 subjects, of whom two subjects dropped out, leaving 14 subjects who completed the USUAL treatment program. Two SAFETY groups initially comprised 20 subjects, of whom four subjects dropped out, leaving 16 subjects who completed the SAFETY group treatment program. The six subjects who dropped out did so within the first few days of treatment and were not included in the data analysis. Overall, 30 subjects completed treatment.
Assessment instruments
The Social Phobia and Anxiety Inventory (SPAI) [15], the Fear of Negative Evaluation Scale (FNE) [16], and the anxiety dimension (anx) of the Symptom Checklist 90-Revised (SCL-90-R) [17] were administered at the beginning and end of treatment. Additionally, subjects rated their subjective anxiety on the Subjective Units of Distress Scale (SUDS) before and after exposure [18]. The SUDS ratings prior to a speech task during a behavioural test on day 3 and day 9 were compared and used as a measure of outcome. The SPAI and FNE were considered to be specific measures for social phobia whilst the SCL-90-R(anx) and SUDS were general anxiety measures.
To monitor treatment integrity, before exposure tasks subjects in the SAFETY group were asked to record the safety behaviours they would normally use in that situation. Following the exposure task, they were asked to record ‘yes’ or ‘no’ as to whether they had engaged in the predicted safety behaviours. Subjects in the USUAL group were asked to record methods of anxiety control they might use in the situation. They too were asked to record whether they had engaged in those behaviours. The behaviours recorded by the SAFETY and USUAL subjects prior to exposure tasks on days 2, 3 and 9 were then blindly assessed by two independent clinicians to determine which behaviours recorded by both groups could be labelled as safety behaviours, based on the Social Behaviours Questionnaire of Clark et al [19]. The chance corrected agreement between clinicians was calculated (see Results).
Procedure
The treatment program extended over 3 weeks. The first week involved full-time attendance at the clinic (days 1–5), the second week involved unsuper-vised practice of exposure and cognitive techniques at home or at work, while the third week (days 6–10) was again spent full-time at the clinic, giving a total of 10 days (80 h) of group treatment. Two clinicians, both of whom were experienced in CBT group treatment of social phobia, ran one SAFETY group and one USUAL group each.
Subjects in the USUAL treatment program received standard CBT as per the Social Phobia Treatment Manual [2]. The SAFETY groups also received standard CBT as per the manual but in addition were educated about the importance of dropping safety behaviours. A clinician guide was devised and used in conjunction with a published Social Phobia Treatment Manual [2] to ensure a standardised treatment approach. The treatment manual comprised; psycho-education about the nature of anxiety and social phobia, instruction on breathing control and relaxation, cognitive therapy, graded exposure, and assertiveness.
As much as possible, the cognitive and behavioural elements were integrated. Both group and individual exposure tasks were utilised to ensure that the idiosyncratic fears of individual subjects were targeted. As per Wells et al. [4], a decrease condition (instruction to drop safety behaviour) was used for the SAFETY group and a neutral condition (instruction to stay in the exposure situation) was used for the USUAL group. Prior to each exposure task, subjects were reminded of these instructions.
Graded practice for a speech task began on day 2. To begin with, subjects read aloud from the treatment manual, then performed impromptu speeches of 15, 30 and 60 s duration in the treatment room at the clinic. A behavioural test (speech task) was administered on the third day of treatment in a small auditorium at the clinic (day 3) and repeated on the ninth day of treatment in the small auditorium during week 3 (day 9). After day 3, subjects continued to perform speeches, which were video-taped, and viewed by the group on each occasion.
Data analysis
Repeated measures of ANOVA were used in the statistical analysis with a confidence interval of 95%. In addition, effect size (ES) indexes in SD units, which reflect change in questionnaire scores before versus after treatment, were calculated for each group [20],[21]. Effect size indexes can be classified into small, medium and large using Cohen's [21] table of effect size index values (small = 0.20 SD units, medium = 0.50 SD units, and large = 0.80 SD units).
Calculation of ES indexes controls for pretest score variance and creates a standardised measure of change across time [22]. The Quality Assurance Project [23] reported a mean (n = 6) ES of 0.98 SD units relative to no-treatment or wait-list control in studies which used behavioural therapy in the treatment of anxiety disorders (panic disorder, generalised anxiety disorder and undifferentiated anxiety states).
Methods
Of the 30 subjects who completed the study, there were no significant pre-treatment differences between the SAFETY and USUAL groups in terms of subject age, sex ratios, proportion taking antidepressant medication, or duration of social phobia (see Table 2). Parameter estimates showed no significant difference between the USUAL and SAFETY groups at pre-treatment for scores on SPAI difference score t= 1.313, p = 0.200, SPAI agoraphobia subscale t = 0.534, p = 0.598, SPAI social phobia subscale t = 1.053, p = 0.302, FNE, t = 0.374, p = 0.712, SCL-R-90(anx) t = −0.872, p = 0.391 and SUDS (day 3) ratings t = −0.720, p = 0.477 (Table 3).
Subject demographic data of the SAFETY and USUAL groups
Subject mean scores for outcome measures of the SAFETY and USUAL groups
Tests of within-subjects contrasts showed significant improvement over time (pre-post) averaging across the USUAL and SAFETY groups on the following measures: SPAI social phobia subscale df = 1, F = 8.224, p = 0.008, SPAI agoraphobia subscale df = 1, F = 7.737, p = 0.010, FNE, df = 1, F= 14.221, p = 0.001, SCL-R-90(anx) df = 1, F= 5.792, p = 0.023 and SUDS (day 3) ratings df = 1, F = 14.483, p = 0.001. However, tests of within-subjects contrasts for the SPAI difference score were not significant (df = 1, F = 3.778, p = 0.062) over time (pre-post) averaging across the USUAL and SAFETY groups.
Tests of within-subjects contrasts showed a significant interaction or difference between the USUAL and SAFETY groups in the amount of change over time for the SPAI. Overall, the specific measures for social phobia (SPAI and FNE) showed large ES gains for the SAFETY group compared to the USUAL group. However, the ES index of the FNE for the SAFETY group was not significantly different compared to the USUAL group. The SPAI social phobia subscale ES index for the SAFETY group was significantly greater than the ES index gain for the USUAL group. Also, the SPAI difference score ES index was significantly greater for the SAFETY group than for the USUAL group. However, the agoraphobia subscale ES indexes were not significantly different between groups (see Table 4).
Outcome measured by effect size (ES) indexes for the SAFETY and USUAL groups
The non-specific anxiety measures (SCL-90-R[anx] and SUDS) did not show significant differences between the SAFETY and USUAL groups. Similar small ES changes in outcome were shown for both groups with the SCL-90-R(anx). However, the SAFETY group experienced a large ES decrease in anticipatory anxiety (SUDS pre-speech) relative to the USUALgroup's medium ES but this was not significant (see Table 4).
Validation of experimental instructions
The chance corrected interrater reliability of the independent clinicians (who blindly assessed subjects' listed anticipated behaviours prior to exposure tasks on days 2, 3 and 9 to determine which behaviours were safety behaviours) was K = 0.59 [24]. A comparison of the number of behaviours identified as safety behaviours by the clinicians between groups showed that subjects in the SAFETY group identified a significantly greater number of safety behaviours compared to the USUAL group, (day 3, t=3.55, df = 26, p = 0.001 and day 9, t = 3.17, df = 25, p = 0.004, respectively).
Following the speech task on day 3 and day 9, 85% and 67% of the SAFETY group reported having used safety behaviours, respectively. By comparison, 77% and 93% of the USUAL group reported using behaviours labelled as ‘anxiety control strategies’ on day 3 and day 9, respectively. However, the percentage change across day 3 and day 9 was not significant for either group, Pearson χ2 (1, n = 12) = 0.30, p = 0.58 (SAFETY group), Pearson χ2 (1, n = 12) = 3.27, p = 0.07 (USUAL group).
Discussion
The results of the present study showed that the USUAL and SAFETY groups made significant improvement in both the specific measures for social phobia and the general measures of anxiety following cognitive-behavioural therapy group treatment. However, the SAFETY group achieved large effect size (ES) index gains for the specific measures of social phobia (SPAI, and FNE), compared to the USUAL group. The ES indexes of the SPAI were significantly greater for the SAFETY group compared to the USUAL group while the ES index of the FNE was not significantly different. These results suggest that instructions to drop safety behaviours were a useful addition to the standard group CBT treatment for social phobia.
The large effect size (ES) index gains for the SAFETY group on the SPAI and the FNE were not artifacts due to problems with treatment integrity as both groups showed significant improvement pre-post on all measures, (except for the SPAI difference score). In addition, the ES indexes for the general anxiety outcome measure SCL-90-R(anx) for both groups (see Table 4) compared well with ES indexes in outcome from previous social phobia group treatment programs at the Anxiety Disorders Unit [Hunt C: unpublished data]. This finding is supported by the SUDS ES indexes which were large (0.85 SD units) for the SAFETY group and medium (0.57 SD units) for the USUAL group (see Table 4).
Turner et al. [25] suggest that the SPAI difference score provides a purer measure of social phobia. They cite a mean SPAI difference score of 111 as typical of untreated patients with generalised social phobia. The pre-treatment SPAI difference scores for the SAFETY and USUAL groups were thus fairly typical for patients with generalised social phobia (see Table 3). The FNE developed by Watson and Friend [16] in 1969 has a long history of use in assessment of social anxiety in the general population. The scale was not developed specifically for social phobia [26] but has been used extensively in studies of social phobia.
The SPAI difference score ES index was interesting given that the USUAL group showed effectively no change pre- versus post-treatment (ES index = −0.04 SD units). The SPAI is designed to assess somatic, cognitive, avoidance, and escape behaviours [27]. Avoidance and escape behaviour can be viewed as the patient's attempt to seek safety (i.e. safety behaviour [7]). The SPAI might be particularly sensitive to change in safety behaviour, as reflected by the measurement of avoidance and escape behaviours. This postulated sensitivity of the SPAI to changes in safety behaviour may explain why the SAFETY group showed improvement on the SPAI measure and the USUAL group did not.
Many subjects in the SAFETY group who engaged in safety behaviours during the speech task on day 3 still used such behaviours on day 9 (85% and 67%, respectively). Contrary to our predictions, the change in reported use of safety behaviours across day 3 and day 9 was not significant. However, the measure of change was crude as subjects were asked only to record yes or no as to whether they had used safety behaviours during exposure. Thus, it was not known how many and which safety behaviours subjects engaged in during exposure. Clinically, subjects stated they dropped some safety behaviours but held on to others during exposure tasks which might explain the high frequency of use of safety behaviours on day 9. Therefore, subjects in the SAFETY group may have decreased the number of safety behaviours they engaged in across time.
Subjects in the USUAL group actually showed an increase in frequency of use of anxiety management strategies during the speech task (77% on day 3 vs 93% on day 9) that approached significance (p = 0.07), possibly reflecting an increased reliance on perceived coping strategies. It is of interest that more than 40% of these predicted anxiety management strategies on each of day 3 and day 9 for the USUAL group were subsequently rated as safety behaviours by the independent raters, suggesting that patients were interpreting safety behaviours as being adaptive coping strategies.
The main limitation of the study was the relative lack of statistical power due to the small sample size. On this small sample, however, repeated measures of ANOVA examining tests of within-subjects contrasts of a specific social phobia measure (SPAI) showed a significant difference between groups. This supports the hypothesis that instructing subjects to drop safety behaviours enhances the efficacy of standard group CBT treatment for social phobia. The results lend support to Clark and Wells' [6] cognitive model of social phobia which postulates that anxiety and negative beliefs about social situations are in part maintained by engaging in safety behaviours. Therefore, the addition of instruction to drop safety behaviours in exposure tasks in the course of CBT treatment for social phobia may be useful. Further replication with larger samples is required and warranted on the basis of the present study.
Future studies could be improved by utilising a standardised measure of safety behaviours pre- and post-treatment for treatment and control. One of the difficulties is that safety behaviours encompass anything the patient does in the belief that it will prevent a feared social danger. This means that a behaviour may be a safety behaviour for one patient but not for another patient. Additionally, future studies could evaluate ways of enhancing subjects' reduction of safety behaviours, such as the insession safety behaviours manipulation described by Clarke [6].
Footnotes
Acknowledgements
The authors wish to thank Ian Hickie, Dusan Hadzi-Pavlovic, David Clark, Heidi Sumich, Lorna Peters, and Gavin Andrews for their valuable contributions to the study.