Introduction
A recent randomized controlled trial, comparing cognitive behavioral group therapy (CBGT), psychodrama (PD), and their integration (CBPT) for patients with social anxiety disorder (SAD), found that all treatments were superior to the waiting list group (WL) in reducing social anxiety symptoms (Abeditehrani et al., 2023). Although processes that affect treatment outcomes in CBT are well-documented (e.g., Hofmann, 2000; Hofmann, 2004), less is known about potential mediators of treatment effect in experiential therapies such as PD and CBPT. Investigating mediators of treatment effect is important to understand both the psychopathological underpinnings of SAD and the therapeutic mechanisms of therapies. Such insights could ultimately lead to the development of more effective treatments (Kraemer et al., 2002).
Cognitive behavior therapies (CBT) for SAD are based on cognitive models that propose that cognitive biases in social situations play an important role in maintaining social anxiety, and that CBT helps by reducing these biases (Clark & Wells, 1995; Heimberg et al., 2010; Hofmann, 2007; Rapee & Heimberg, 1997). Particularly, individuals with SAD tend to overestimate the likelihood of negative outcomes in social situations and exaggerate the consequences of these outcomes (Voncken et al., 2003; Foa et al., 1996; Lucock et al., 1988; McManus et al., 2000). This maladaptive thinking leads to increased anxiety, prompting anxious individuals to avoid social situations to prevent this anxiety (Mesa et al., 2011). Several studies showed that, indeed, CBT effectively reduces social anxiety symptoms in individuals with SAD by targeting and modifying these cognitive biases in both probability and cost estimates (Foa et al., 1996; McManus et al., 2000; Gregory et al., 2015; Hofmann, 2004).
Although, to our knowledge, this has never been studied, experiential therapies like PD also could also be effective because they reduce cognitive biases. Experiential therapies that involve enacting anxiety-provoking social situations allow socially anxious patients to test and correct their negative cognitions through the enactment of anxiety-provoking social situations. Several acting techniques in psychodrama (see Abeditehrani et al., 2020 for a description of typical psychodrama techniques and their goals for treatment of SAD), help patients change their beliefs and attitudes about the anxiety provoking social situations (Treadwell, & Kumar, 2002). For example, an often-used experiential technique in PD is role reversal. In a role reversal, two individuals first engage in role-playing. Then, they switch positions and take on each other’s role (Moreno, J. L, Moreno, Z., & Moreno. J., 1995). This technique challenge and potentially correct the negative self-perceptions that often lead individuals with social anxiety to overestimate the likelihood of negative judgment (Clark & Wells, 1995; Rapee & Heimberg, 1997). A recent experiment indicated that role reversal is an effective technique that can be used to correct such negative cognitions, reducing the estimated cost and probability of being judged negatively (Abeditehrani et al., 2021). These findings suggest that the modifying cognitive biases is not unique to cognitive restructuring in CBGT but can also occur during experiential techniques in PD.
Although research on avoidance as a mechanism of change in CBT is limited, it is frequently proposed as a key factor due to the exposure-based nature of these interventions (Goldin et al., 2016; Teachmen, Beadel & Steinman, 2014). Exposure-based interventions highlight that avoidance behaviors are essential targets in therapy as they prevent disconfirmation of negative beliefs and may make some feared outcomes more likely (for a comprehensive review, Moscovitch et al., 2009). Evidence suggests that successfully reducing avoidance enables patients to acquire new information regarding the likelihood of feared consequences and to process new emotional information that contradicts their existing fear-related mental representations, thereby reducing the fear (Foa & Kozak, 1986; Foa & McNally, 1996; Moscovitch et al., 2009). Cognitive behavioral therapy of SAD includes roleplayed exposure in sessions and in vivo exposure to feared situations to reduce avoidance as homework (Clark & Wells, 1995; Rapee & Heimberg, 1997). In experiential interventions, role playing during the enactment of a person’s core concerns is used to confront patients with feared situations and to decrease avoidance (Blatner, 1996). Thus, PD might also be effective by reducing avoidance behavior.
In addition to shared mediators of treatment outcome, experiential therapies may also operate through processes specific to these types of interventions. Psychodrama theory proposes that a lack of spontaneity plays an important role in maintaining anxiety (Moreno, 1946). Lack of spontaneity refers to an individual’s difficulty in generating novel, effective, or adaptive responses to new or challenging situations. According to Moreno’s theory, increasing spontaneity will decrease anxiety (Moreno, 1946). Moreno emphasized the importance of enhancing spontaneity to enable individuals to respond to situations without inhibition (Moreno, 1953; Blatner, 2000; Collins et al., 1997). Indeed, inhibition marked by a persistent tendency to react cautiously or avoid unfamiliar people, places, and objects, is one of the most recognized risk factors for SAD (Kagan et al., 1988). Spontaneity, as the opposite of inhibition should thus reduce social anxiety. In psychodrama (PD) (experiential) techniques are used to encourage spontaneity, which is hypothesized to be a central mechanism of change in this treatment (Kellerman, 1992). Despite its theoretical importance, we are not aware of any research examining spontaneity as a mediating factor in psychodrama therapy.
Recently, a new intervention was created that integrated CBGT and PD (labelled Cognitive Behavioral Psychodrama Therapy; CBPT) (Abeditehrani et al., 2020). This intervention integrated cognitive restructuring into the enactment of anxiety-provoking social situations and may therefore be effective through both shared mechanisms (such as reducing avoidance and modifying cognitions) and its impact on spontaneity. To summarize, we examine whether, in CBGT, treatment effects are mediated by reductions in avoidance and changes in cognitive biases, specifically in the perceived cost and probability of negative social events. For PD and integration of CBGT and PD (labeled CBPT), we expect that these treatments not only reduce avoidance and alter cognitive biases but also foster spontaneity, which serves as an additional mediator.
Method
A randomized controlled trial (RCT) was conducted comparing CBGT, PD, and CBPT as treatments for SAD to compare active treatment with WL. The data presented in this paper are from this RCT (see Abeditehrani et al., 2023). The study was registered at the Iranian Registry of Clinical Trial (IRCT2016032321385N1), and an Iranian ethical committee (reference number IR.UMSHA. REC.1394.521) approved the protocol on February 27, 2016. Random allocation to each condition was done by an independent staff member of the University of Amsterdam using a research randomizer (https://www.randomizer.org/). The detailed description and Consort diagram of these treatments are described elsewhere (see Abeditehrani et al., 2023).
Participants
The study included 116 Iranian patients for whom complete baseline, mid-, and post-treatment data were available. The inclusion criteria primary diagnosis of SAD, age between 18 and 65 years, ability to read and understand the questionnaires and the interview. Potential participants were excluded if they had comorbid psychotic or bipolar disorder, a high suicidality risk, a comorbid diagnosis of substance abuse or dependence, or an unwillingness to stabilize medication for the duration of the study (nine months). Their age ranged from 18.06 to 65.89 years, M = 28.21, SD = 7.44. Twenty-nine received CBGT, 24 PD, 27 CBPT, and 36 were on the waitlist. Most participants were female (67%), held a university level education (69%) and were unemployed (71.6%).
Measures
Primary Outcome
The Liebowitz Social Anxiety Scale – clinician-administered version (LSAS; Liebowitz, 1987) is a 24-item interview that assesses fear and avoidance on a 4-point-Likert scale (0 = never to 3 = usually). The scale contains items about social interactions (e.g., talking with people you don’t know very well), and performance situations (e.g., returning items to a store). The LSAS has shown strong test–retest reliability, internal consistency, and valid convergence and discrimination (Baker et al., 2002; Fresco et al., 2001; Oakman et al., 2003; Rytwinski et al., 2009). In analyses where avoidance was tested as mediator, the avoidance scores of LSAS were excluded from the total LSAS score to prevent conceptual overlap. The remaining fear subscale of the LSAS showed a Cronbach’s alpha of 0.95, indicating sufficient internal consistency for this outcome measure.
Mediators
The Outcome Probability Questionnaire (OPQ) and the Outcome Cost Questionnaire (OCQ), developed by Uren et al. (2004), are two self-report instruments comprising 12 items each. The OPQ assesses individual’s likelihood estimations of experiencing negative social events within the next year (e.g., at a party, others will notice that you are nervous). The OPQ uses a 9-point Likert scale ranging from 0 = not at all likely to 8 = extremely likely. In addition, the OCQ assesses the perceived cost if these events were indeed to occur by asking participants to rate how bad or distressing the outcomes would be for them. This questionnaire uses a 9-point Likert scale ranging from 0 = distressing to 8 = extremely distressing. Both instruments have shown good internal consistency, with Cronbach’s alpha values exceeding 0.90 (Uren et al., 2004).
The Personal Attitude Scale-II (PAS), developed by Kellar et al. (2002), is a self-report measure assessing spontaneity. An example item is: “I can easily accept when plans change quickly.”. The scale comprises of 66 items rated on a 5-point Likert scale (ranging from 0 = strongly disagree to 4 = strongly agree). The questionnaire showed a Cronbach’s alpha reliability coefficient of 0.92 for internal consistency and a test-retest reliability of 0.86 (Kellar et al., 2002).
The Social Avoidance and Distress Scale (SADS) is a self-report inventory developed by Watson and Friend (1969) with 28 items that are evenly split into two subscales, one for assessing social avoidance (e.g., I often want to get away from people), and another for assessing social anxiety (e.g., I often feel on edge when I am with a group of people). Participants answer each item with ‘true’ or ‘false’. Watson and Friend (1969) reported a Cronbach’s alpha reliability coefficient of 0.90 and a test-retest reliability of 0.77 for the scale. Due to substantial conceptual overlap between the distress scale of the SADS and the fear subscale of the LSAS, only the avoidance subscale of the SADS was used in the analysis in which avoidance was regarded as mediator. This subscale showed a coefficient Cronbach’s alpha reliability of 0.82 in the present sample.
Interventions
Participants receiving treatment underwent a 12-week group intervention, with each weekly session lasting 2.5 h. Each group consisted of six patients and was led by two therapists (one male and one female), who were trained and experienced in both CBGT and PD. To ensure the integrity of the study and avoid contamination between the interventions, therapists received weekly group supervision from an expert. Participants who were placed on a waiting list for 12 weeks received no active intervention and were assessed at the same time points as the treatment groups to ensure comparability in data collection.
CBGT
This study used Heimberg and Becker’s (2002) CBGT protocol for Social Anxiety Disorder (SAD), with initial sessions focusing on presenting the cognitive-behavioral model of social anxiety and provide training in cognitive restructuring. Sessions 3 through 11 concentrate on practicing exposure alongside cognitive restructuring. Homework assignments involve real-life exposure situations. The therapy concludes with a session that reviews patient progress. In this session, therapists assist patients in setting goals for continued practice in anxiety-provoking situations after the ending of treatment.
PD
The first psychodrama session introduces participants to the principles and techniques of this experiential approach (Moreno, 1946). All other sessions follow a three-stage structure. Each session begins with verbal or non-verbal warm-up techniques (see Weiner & Sacks, 1969). Next, a patient is selected to be the protagonist, the main character of the action stage. During the action stage, the protagonist’s anxiety-provoking situation is acted out using various psychodrama techniques, with the help of other patients who act as auxiliary ego(s). For a detailed description of these techniques readers are referred to Abeditehrani et al. (2020). The sessions conclude with a sharing stage, where patients express their feelings and thoughts. In the last session, the sharing stage involves reflections on all previous sessions and the patient’s progress throughout the treatment. Unlike CBGT, PD does not include homework assignments.
CBPT
The initial two sessions of the integrated CBPT are similar to CBGT, including cognitive restructuring training and assigned homework. Sessions 3 to 11 follow the stages of psychodrama (PD), but here therapists integrate CBGT and PD techniques. The warm-up stage is like psychodrama warm-up, but during the action stage, therapists incorporate CBGT techniques, specifically cognitive restructuring, into the PD enactment. There are multiple ways to integrate CBGT and PD techniques during the action stage. One approach is for therapists to apply specific psychodrama techniques that support cognitive restructuring during the enactment. For example, they may use the “role reversal” technique, where the protagonist switches roles with another person in the scenario (e.g., an authority figure or peer). In CBPT, the therapist may prompt the protagonist to engage in cognitive restructuring while in the other person’s role, which can help the protagonist reframe their original thoughts and emotions from a new perspective. Another method is for therapists to incorporate cognitive restructuring directly into the PD enactment. They pause the scene at key moments to identify negative thoughts and guide the protagonist through cognitive restructuring before resuming with new perspectives. Additionally, therapists might have the protagonist first enact the situation with the help of auxiliary egos, then the therapist can guide the protagonist to identify their negative automatic thoughts and challenge them through Socratic questioning, while the group observes the process to engage in cognitive restructuring, and finally reenact the situation using alternative, more adaptive thoughts. Like PD, the last part of each session provides an opportunity for participants to share their feelings and thoughts. Only the session’s protagonist is assigned a homework task, which involves exposure in vivo. Following the CBGT protocol, the last session is a conclusion session. For a comprehensive description of the techniques and objectives of this treatment see Abeditehrani et al. (2020).
Procedure and Design
After recruitment, patients were diagnosed with the structured clinical interview for DSM 4th edition (SCID-I) (First et al., 2012; Sharifi et al., 2007). The RCT study conducted between October 2015 and June 2018 in Iran. To assess the diagnosis of SAD, the SCID-I was administered prior to inclusion. The assessor was PhD student in clinical psychology who had received a SCID training. Patients were assessed with the outcome measure and the mediators. For the current analysis we use the assessments at baseline, middle (after 6th session), and after the last session. Each assessment included an LSAS interview that was conducted in the clinic by an independent assessor who was blind to the patient’s treatment-condition. The other questionnaires were assessed online at home by sending a link to the questionnaires right after the LSAS interview.
Analysis
To test the hypotheses, we conducted separate mediation analyses per condition (CBGT, PD, CBPT) and used the WL as a reference group. Modeling random effects for treatment groups led to estimation failure; therefore, such effects could not be incorporated into the model. The outcome measure was the LSAS. Recent research showed that mediation analysis, with baseline scores as covariates, has good performance in testing possible mediators in pretest-posttest control group designs (Valente & MacKinnon, 2017). Ideally, the mediator is assessed before the outcome. Mid-treatment assessments seem suitable for use as mediator variables within the present study. However, the drawback of this is that particularly in PD and CBPT, there are substantial differences between patients, with some selected as protagonists in the first half of treatment and others in the second half. Thus, to capture the full effect of PD and CBPT on the proposed mediator variables, we also included post-assessment scores for exploratory analysis. The limitation of using post-assessment scores is that the mediator was not assessed before the outcome, which affects the causal interpretation. Note that the waitlist period did not extend until the follow-up measure, which occurred six months later, hence follow-up couldn’t be used as outcome measures. We therefore decided to run both mediation analyses, one with the mediator assessed mid-treatment, and one with the mediator assessed at post-treatment. We utilized post-assessment LSAS scores as the dependent variable, mid-treatment, respectively post-measurement as mediator, and baseline measurements (both outcome baseline and mediator baseline) as covariates. For mediation analysis, SPSS PROCESS (Hayes, 2017) model 4 with 5,000 bootstrap samples was conducted with data from cases without missing values. There were a few cases with missing data: PD = 1, CBPT = 2, and WL = 1. Mediation effects were tested by evaluating the upper and lower boundaries of the 95% bias-corrected bootstrap confidence intervals (95% BC CI) of the indirect effect.
Results
For descriptive statistics, see Table 1. Results of the mediation analyses for mediators assessed mid-treatment are reported in Table 2 and mediators assessed post-treatment are reported in Table 3.
Table 1
Baseline, Middle, and Posttreatment estimated means (and Standard Errors) of CBGT, PD, CBPT, and WL Condition for all outcomes and proposed mediation variables
Condition | Baseline | Middle | Post-assessment | |
|---|---|---|---|---|
Outcomes, M (SE) | ||||
LSAS | CBGT | 80.62 (4.31) | 57.72 (4.44) | 39.58 (4.03) |
PD | 89.29 (4.74) | 60.57 (4.21) | 44.41 (4.52) | |
CBPT | 83.37 (4.47) | 57.52 (4.69) | 36.36 (4.33) | |
WL | 83.44 (3.87) | 78.74 (3.31) | 77.04 (3.66) | |
Mediators, M (SE) | ||||
Outcome cost | CBGT | 67.93 (3.36) | 58.07 (3.92) | 43.28 (3.89) |
PD | 73.33 (3.84) | 63.43 (4.09) | 49.74 (4.42) | |
CBPT | 76.33 (3.17) | 55.20 (4.51) | 45.92 (3.98) | |
WL | 73.86 (3.03) | 69.17 (2.89) | 70.94 (3.14) | |
Outcome probability | CBGT | 57.97 (3.19) | 50.34 (3.25) | 34.97 (3.21) |
PD | 57.67 (3.92) | 47.91 (4.21) | 39.78 (4.33) | |
CBPT | 53.74 (3.94) | 45.48 (3.74) | 39.72 (4.26) | |
WL | 55.47 (2.84) | 55.77 (2.71) | 57.00 (2.93) | |
Spontaneity | CBGT | 111.90 (5.43) | 120.07 (5.46) | 142.17 (5.56) |
PD | 103.33 (4.41) | 116.61 (5.73) | 127.43 (5.92) | |
CBPT | 108.59 (3.62) | 124.04 (4.11) | 142.40 (5.40) | |
WL | 103.92 (4.56) | 107.06 (5.14) | 101.91 (4.57) | |
Avoidance | CBGT | 18.62 (1.15) | 14.31 (1.14) | 9.07 (1.17) |
PD | 20.88 (0.90) | 15.70 (1.40) | 12.00 (1.54) | |
CBPT | 18.59 (1.18) | 13.24 (1.26) | 8.80 (0.99) | |
WL | 19.69 (0.90) | 19.03 (1.02) | 18.83 (1.05) |
Table 2
Mid-treatment mediation analysis for various treatment conditions (CBGT, PD, CBPT, with WL as a reference group)
Condition | Mediator | Dependent Variable | Effect of IV on M (a) | Effect of M on DV (b) | Indirect effect (ab) | Direct effect (c’) | Total effect (c) |
|---|---|---|---|---|---|---|---|
CBGT vs. WL | Outcome Cost | LSAS | b = -10.13 (4.38), t = -2.31* | b = 0.49 (0.13), t = 3.80*** | b = -4.99, 95% BC CI [-10.63, -0.22] | b = -31.42 (4.59), t = -6.85*** | b = -36.42 (4.87), t = -7.48*** |
Outcome Probability | LSAS | b = -5.24 (3.84), t = -1.37 | b = 0.39 (0.16), t = 2.49* | b = -2.05, 95% BC CI [-7.94, 0.61] | b = -33.45 (4.74), t = -7.06*** | b = -35.51 (4.87), t = -7.29*** | |
Spontaneity | LSAS | b = 5.97 (5.04), t = 1.18 | b = -0.28 (0.12), t = -2.40* | b = -1.65, 95% BC CI [-7.32, 0.69] | b = -32.57 (4.56), t = -7.15*** | b = -34.22 (4.68), t = -7.31*** | |
Avoidance | LSAS Fear | b = -1.54 (0.64), t = -2.41* | b = 1.87 (0.47), t = 3.98*** | b = -2.88, 95% BC CI [-6.80, -0.17] | b = -13.60 (2.43), t = -5.59*** | b = -16.48 (2.60), t = -6.35*** | |
PD vs. WL | Outcome Cost | LSAS | b = -7.09 (4.04), t = -1.75 | b = 0.37 (0.16), t = 2.31* | b = -2.65, 95% BC CI [-7.01, 0.92] | b = -32.81 (4.94), t = -6.64*** | b = -35.45 (4.99), t = -7.10*** |
Outcome Probability | LSAS | b = -9.00 (4.27), t = -2.11* | b = 0.58 (0.14), t = 4.13*** | b = -5.26, 95% BC CI [-12.31, -0.37] | b = -29.61 (4.62), t = -6.40*** | b = -34.87 (5.06), t = -6.89*** | |
Spontaneity | LSAS | b = 8.74 (5.86), t = 1.49 | b = -0.33 (0.11), t = -2.98** | b = -2.87, 95% BC CI [-9.93, 0.64] | b = -31.98 (4.84), t = -6.61*** | b = -34.85 (5.08), t = -6.86*** | |
Avoidance | LSAS Fear | b = -1.87 (0.71), t = -2.64* | b = 1.81 (0.42), t = 4.29*** | b = -3.39, 95% BC CI [-6.75, -0.68] | b = -13.72 (2.34), t = -5.87*** | b = -17.11 (2.53), t = -6.76*** | |
CBPT vs. WL | Outcome Cost | LSAS | b = -13.32 (4.84), t = -2.75** | b = 0.31 (0.13), t = 2.33* | b = -4.12, 95% BC CI [-10.22, -0.12] | b = -35.98 (5.13), t = -7.02*** | b = -40.10 (5.00), t = -8.02*** |
Outcome Probability | LSAS | b = -9.29 (4.03), t = -2.30* | b = 0.55 (0.16), t = 3.54*** | b = -5.11, 95% BC CI [-11.77, -0.57] | b = -35.43 (4.91), t = -7.22 *** | b = -40.54 (5.15), t = -7.87*** | |
Spontaneity | LSAS | b = 12.90 (5.52), t = 2.34* | b = -0.18 (0.12), t = -1.46 | b = -2.30, 95% BC CI [-8.35, 0.30] | b = -37.63 (5.31), t = -7.09*** | b = -39.93 (5.12), t = -7.80*** | |
Avoidance | LSAS Fear | b = -2.15 (0.70), t = -3.08** | b = 1.37 (0.49), t = 2.81** | b = -2.94, 95% BC CI [-6.70, -0.38] | b = -16.63 (2.75), t = -6.05*** | b = -19.57 (2.69), t = -7.26*** |
Table 3
Post-treatment mediation analysis for various treatment conditions (CBGT, PD, CBPT, with WL as a reference group)
Condition | Mediator | Dependent variable | Effect of IV on M (a) | Effect of M on DV (b) | Indirect effect (ab) | Direct effect (c’) | Total effect (c) |
|---|---|---|---|---|---|---|---|
CBGT vs. WL | Outcome Cost | LSAS | b = -26.22 (4.94), t = -5.31*** | b = 0.54 (0.11), t = 5.08*** | b = -14.24, 95% BC CI [-26.11, -5.68] | b = -22.18 (4.96), t = -4.47*** | b = -35.42 (4.87), t = -7.48*** |
Outcome Probability | LSAS | b = -21.73 (4.11), t = -5.29*** | b = 0.55 (0.14), t = 3.99*** | b = -11.87, 95% BC CI [-24.34, -3.30] | b = -23.64 (5.28), t = -4.48 *** | b = -35.51 (4.87), t = -7.29*** | |
Spontaneity | LSAS | b = 33.66 (4.95), t = 6.80*** | b = -0.42 (0.11), t = -3.79*** | b = -14.08, 95% BC CI [-24.15, -5.21] | b = -20.14 (5.63), t = -3.57*** | b = -34.22 (4.68), t = -7.31*** | |
Avoidance | LSAS Fear | b = -4.39 (0.77), t = -5.73*** | b = 1.29 (0.41), t = 3.17** | b = -5.66, 95% BC CI [-10.27, -1.61] | b = -10.81(3.01), t = -3.59 *** | b = -16.48 (2.60), t = -6.35 *** | |
PD vs. WL | Outcome Cost | LSAS | b = -21.64 (5.27), t = -4.11*** | b = 0.65 (0.09), t = 6.91*** | b = -14.12, 95% BC CI [-24.12, -6.21] | b = -21.33 (4.19), t = -5.09*** | b = -35.45 (4.99), t = -7.10*** |
Outcome Probability | LSAS | b = -18.11 (4.69), t = -3.86*** | b = 0.61 (0.12), t = 4.97*** | b = -11.02, 95% BC CI [-23.81, -3.56] | b = -23.85 (4.77), t = -5.00*** | b = -34.87 (5.06), t = -6.89*** | |
Spontaneity | LSAS | b = 25.09 (6.05), t = 4.14*** | b = -0.55 (0.09), t = -6.35*** | b = -13.83, 95% BC CI [-22.84, -5.79] | b = -21.02 (4.43), t = -4.74*** | b = -34.85 (5.08), t = -6.86*** | |
Avoidance | LSAS Fear | b = -4.08 (0.84), t = -4.83*** | b = 1.75 (0.33), t = 5.24 *** | b = -7.15, 95% BC CI [-12.32, -2.72] | b = -9.96 (2.48), t = -4.01*** | b = -17.11 (2.53), t = -6.76*** | |
CBPT vs. WL | Outcome Cost | LSAS | b = -24.77 (4.98), t = 4.98*** | b = 0.53 (0.12), t = 4.57 *** | b = -13.06, 95% BC CI [-22.51, -6.26] | b = -27.04 (5.16), t = -5.24*** | b = -40.10 (5.00), t = -8.02*** |
Outcome Probability | LSAS | b = -16.40 (4.69), t = -3.49*** | b = 0.41 (0.14), t = 2.99** | b = -6.73, 95% BC CI [-17.73, -1.24] | b = -33.81 (5.32), t = -6.36 *** | b = -40.54 (5.15), t = -7.87*** | |
Spontaneity | LSAS | b = 35.74 (5.00), t = 7.15*** | b = -0.41 (0.13), t = -3.22** | b = -14.57, 95% BC CI [-26.79, -5.34] | b = -25.36 (6.55), t = -3.87*** | b = -39.93 (5.12), t = -7.80*** | |
Avoidance | LSAS Fear | b = -5.08 (0.60), t = -8.47*** | b = 1.91(0.55), t = 3.48 ** | b = -9.70, 95% BC CI [-16.16, -3.84] | b = -9.87 (3.72), t = -2.65* | b = -19.57 (2.69), t = -7.26*** |
Outcome cost
As can be seen in Table 2; Fig. 1, expected outcome cost mediated treatment effects CBGT and CBPT when the mediator was assessed mid-treatment. For PD, however, the results indicated that outcome cost did not mediate the treatment effect, contrary to our hypotheses. The results of the analysis with post-assessment as a mediator for perceived cost of negative social outcomes are displayed in Table 3. Outcome cost served as a possible mediator CBGT, PD, and the integrated treatment CBPT.
Fig. 1
Graphical representation of the multiple post-treatment (left) / mid-treatment (right) mediation effects for outcome cost for CBGT, PD, and CBPT vs. WL. Note a = effect of condition on mediator; b = effect of mediator on outcome; c = total effect of condition on outcome (without the influence of the mediator in the model); c’ = effect of condition on outcome when controlled for mediator in the mediation model; BC CI = lower and upper level of 95% bootstrap-corrected confidence interval (based on 5000 samples). *P < .05, **P < .01, ***P < .001
×
![]()
Outcome Probability
The results of the analysis with mid-assessment mediator for probability are shown in Table 2; Fig. 2. Outcome probability assessed at mid-treatment mediated the treatment effects of PD and integrated CBPT, but not of CBGT. For CBGT, this result did not align with our hypotheses. For all treatments, outcome probability served as a possible mediator in post-treatment assessments (see Table 3).
Fig. 2
Graphical representation of the multiple mid-treatment mediation effects for outcome probability for CBGT, PD, and CBPT vs. WL. Note a = effect of condition on mediator; b = effect of mediator on outcome; c = total effect of condition on outcome (without the influence of the mediator in the model); c’ = effect of condition on outcome when controlled for mediator in the mediation model; BC CI = lower and upper level of 95% bootstrap-corrected confidence interval (based on 5000 samples). *P < .05, **P < .01, ***P < .001
×
![]()
Spontaneity
The results of the analysis with mid-assessment spontaneity as mediator are displayed in Table 2; Fig. 3. Mid-treatment spontaneity did not mediate treatment outcomes in any active interventions comparing to WL. For the post-treatment outcomes, spontaneity served as a possible mediator in both PD and the integrated CBPT. Contrary to the hypotheses, spontaneity also served a possible mediator in CBGT (see Table 3).
Fig. 3
Graphical Representation of the Multiple Post-treatment (left) / Mid-treatment (right) Mediation Effects for Spontaneity for CBGT, PD, and CBPT vs. WL. Note a = effect of condition on mediator; b = effect of mediator on outcome; c = total effect of condition on outcome (without the influence of the mediator in the model); c’ = effect of condition on outcome when controlled for mediator in the mediation model; BC CI = lower and upper level of 95% bootstrap-corrected confidence interval (based on 5000 samples). *P < .05, **P < .01, ***P < .001
×
![]()
Avoidance
Investigating the impact of avoidance, our analysis showed that, as hypothesized, mid-treatment avoidance mediated the effect of all treatments on LSAS, see Table 2; Fig. 4. Moreover, post-treatment avoidance was a significant possible mediator of treatment effects of all interventions compared to WL (see Table 3).
Fig. 4
Graphical Representation of the Multiple Mid-treatment Mediation Effects for Avoidance for CBGT, PD, and CBPT vs. WL. Note a = effect of condition on mediator; b = effect of mediator on outcome; c = total effect of condition on outcome (without the influence of the mediator in the model); c’ = effect of condition on outcome when controlled for mediator in the mediation model; BC CI = lower and upper level of 95% bootstrap-corrected confidence interval (based on 5000 samples). *P < .05, **P < .01, ***P < .001
×
![]()
Discussion
This study aimed to determine mediators of treatment effect of PD and CBGT and their integration CBPT. Specifically, we investigated if the effect of these interventions is mediated by changes in avoidance and cognitive biases such as perceived likelihood and anticipated cost of negative social events, while for PD and CBPT, spontaneity was expected to be an additional mediator. Both mid- and post-treatment measures were used as mediators in the analyses as each approach have its advantages and disadvantages. Results indicated that, while avoidance mediated treatment effects across all interventions, treatment specific to mediators related to cognitive biases were found. Furthermore, results showed that spontaneity when assessed mid-treatment was not a mediator, and when assessed post-treatment cannot be considered to be a unique mediator of treatment effects for experiential therapies.
Avoidance behavior emerged as a significant mediator across all treatment modalities, both mid-treatment and post-treatment, reinforcing the importance of addressing avoidance in treating SAD. All treatments involved exposure to feared social situations, either through implementing within-session and in vivo exposure (CBGT) or through enactment (PD and CBPT). The finding that avoidance behavior mediated the treatment effect in CBGT is consistent with existing literature (e.g., Goldin et al., 2016).
Changes in cognitive biases, specifically the perceived cost and/or probability of negative social outcomes, were significant mediators of all treatment outcomes, with nuanced differences between treatments. As it was expected in context of PD, cognitions also played a mediating role supporting the idea that experiential techniques lead to corrections in biases. CBPT showed clear mediation effects for both biases at mid-treatment and post-treatment, while CBGT and PD had mixed effects. In CBGT, changes in perceived cost, but not probability, served as a mediator, suggesting that cognitive interventions work through cost estimates, consistent with previous research (Hofmann, 2004; O’Toole et al., 2014). Furthermore, Hofmann’s research (2007) demonstrated that using direct cognitive interventions led to better maintenance of treatment benefits, by mediating reductions in perceived social cost during therapy. Conversely, in PD, only the reduction in perceived likelihood of negative social outcomes at mid-treatment mediated outcomes. This effect was likely due to early role-playing enactments that allowed patients to test the likelihood of negative outcomes. These (non-hypothesized) findings, suggest important differences in how each treatment achieves its outcome. It also suggests that the integration might be more robust as it affected both mediators. Further research is needed to determine what aspects of these techniques are responsible for the differences in mechanisms. It should be noted however that the specificity was only observed with the mid-treatment mediators.
The role of spontaneity as a mediator was inconclusive, with different effects observed at mid-treatment and post-treatment. Contrary to our initial hypothesis, spontaneity did not mediate treatment effects at mid-treatment in any intervention. However, by post-treatment, spontaneity served as a possible mediator for all treatments, including CBGT. A common experiential technique to boost spontaneity, central to Moreno’s psychodrama theory, is role-playing. This technique is rooted in psychodrama and behavioral therapy as a learning tool (Moreno, 1946; Kelly, 1955; Lazarus, 1965; Wolpe, 1958). In CBGT, role-playing is often used as an exposure technique, but it may also enhance spontaneity in individuals with SAD (Heimberg & Becker, 2002; Wells & McMillan, 2004). However, without temporal precedence, it is difficult to claim that spontaneity fully mediates treatment effects. It is equally plausible that reductions in anxiety facilitated improvements in spontaneity, rather than spontaneity driving the reduction in anxiety. These findings emphasize the need for cautious interpretation, particularly when considering the role of post-treatment variables, which cannot definitively establish mediation but instead suggest exploratory insights into possible mediators at play. This might also suggest that improvement in spontaneity requires more time to develop over the course of treatment before it acquires a mediating role.
Given these findings, it is important to acknowledge both the implications and the limitations of the current study. One limitation is that participants in the waiting list condition were not included in the follow-up measures, which means we cannot determine the durability of the mediating effects over time. Future research should include long-term follow-up assessments to further determine these mediating mechanisms. Additionally, the small sample size in the PD condition (n = 24) may have limited our ability to detect significant effects, raising the possibility that non-significant findings (e.g., spontaneity) reflect insufficient statistical power rather than a true lack of effect. Moreover, there was some conceptual and empirical overlap among the mediators, with correlations ranging from 0.3 to 0.6. Although these correlations suggest some association, they are not high enough to conclude that they measure the same construct. Due to the sample size, conducting a parallel mediation model to test multiple (correlated) mediators in a single model was not feasible. Instead, we used several separate mediation models with small comparison groups (e.g., 24 vs. 36 WL participants). This approach may have increased the risk of spurious associations, highlighting the importance of replication in future studies. Another limitation is the absence of a formal treatment fidelity assessment by independent judges. Although therapists were fully aware that the groups were part of a research study and were trained to follow the treatment protocols and received weekly supervision to ensure consistency and minimize the risk of contamination between conditions, an independent fidelity check (e.g., video reviews or fidelity checklists) could offer a more objective measure of treatment integrity.
Whereas mid-treatment mediators showed specific effects, analysis using post-treatment mediators were all significant and showed no treatment specificity and indicated that all proposed mediators were associated with treatment outcome. Although the findings of post-treatment mediators should be interpreted cautiously, it is also plausible that mediators of treatment outcome emerge in the second half of the treatment. For example, because half of the participants did not yet be protagonist in the first half of PD and CBPT. It remains uncertain when individuals achieve improvements in specific treatments. This would be an important question for future research, and more frequent assessments would be needed to better understand the timing of these changes and their cumulative impact.
In conclusion, this study highlights the critical importance of addressing avoidance behaviors and cognitive biases in the treatment of social anxiety disorder (SAD). Furthermore, for cognitive biases specificity was found with the integrated treatment showing changes in both cognitive biases as mediators of outcome, suggesting increased robustness. These findings enhance our understanding of the distinct and overlapping mediators in these therapies.
Declarations
Competing Interests
The authors declare no competing interests.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.