Introduction

Over the last decades, psychotherapy research has provided empirical support for psychological treatments. To date, exposure-based treatments have demonstrated some of the largest effects in treating anxiety disorders such as specific phobias (SP) (e.g., Abramowitz et al., 2019; Botella et al., 2016). Meta-analytic findings suggest that repeated systematic engagement with feared stimuli is an essential component of treatments for anxiety disorders (Kaczkurkin & Foa, 2015).

Exposure involves helping patients to confront their feared stimuli on a repeated and prolonged basis (Abramowitz et al., 2019) and can be conducted in different forms: In Vivo Exposure (IVE), imaginal, and through Virtual Reality (VR) (Eaton et al., 2018). In IVE, therapists have their clients confront feared stimuli in a safe, real-life environment (Hazlett-Stevens & Craske, 2009). The real-life exposure to feared stimuli can provide a real confrontation of feared stimuli, and allows patients to learn and tolerate their anxiety at the moment. This form of exposure can be considered the most traditional and has gathered substantial evidence for its effectiveness.

Throughout the years, new ways to conduct exposure have emerged. One way to approach exposure includes the use of technology which has shown to be a viable option for the treatment of SP (Botella et al., 2016, 2017; Maples-Keller et al., 2017). Among technological alternatives we find Augmented Reality Exposure (ARE). This technique combines the real world with virtual elements in real-time. This means that the person sees an image composed of a visualization of the real world and a series of virtual elements, which provides the person with relevant information that is not found in the real world (e.g., the person sees through the device the real environment but virtual cockroaches are added for activating fear).

Research has shown that ARE is an effective treatment for SP such as aviophobia (Rothbaum et al., 2006; Maltby et al., 2002; Shiban et al., 2017), arachnophobia (Garcia-Palacios et al., 2002), and claustrophobia (Botella et al., 2000; Wiederhold & Bouchard, 2014). It has also been demonstrated to be a useful tool in small animal phobias treatment (Botella et al., 2016; Wrzesien et al., 2015). As it gives the patient the opportunity of confronting the phobic stimuli in a real environment by means of VR and consequently extinguished by repeated and controlled exposure (Botella et al., 2017; Emmelkamp & Meyerbröker, 2021; Fernández-Álvarez, et al., 2020).

Both IVE and ARE have shown to be—on average—highly effective (Wechsler et al., 2019). However, there is scarce evidence regarding specific predictors of treatment success. Identifying those specific ingredients that make treatment effective is important to understand the underlying mechanisms of change better. Recent systematic reviews of factors influencing the success of exposure therapy for SP identify elements such as inhibitory learning, low trait anxiety, and high motivation (Böhnlein et al., 2019); affective styles and adjusting style (Totzeck et al., 2019); and patients with poor prognosis (Leehr et al., 2021).

A variable that has been shown to be a robust predictor of treatment outcome is the therapeutic alliance (Flückiger et al., 2018; Pan et al., 2011; Weck et al., 2015). According to the widely accepted tripartite definition proposed by Bordin (1979), this construct includes that: (a) the therapist and patient connect in a mutually supportive and respectful way (bond alliance), (b) the patient and therapist agree on treatment goals (goal alliance), and (c) the patient and therapist agree on the tasks implemented to reach such goals (task alliance). The therapeutic alliance may be critical in exposure-based treatments, as they involve confrontation with feared stimuli instead of avoidance (Buchholz & Abramowitz, 2020). Given this emphasis on behavioral change, research suggests that the task component (in comparison to the bond and goal alliance) has the most profound effect on the reduction of anxiety symptoms (Hagen et al., 2016; Wheaton et al., 2016).

Therapeutic alliance has been associated with treatment outcome in a one-session exposure for SP (Pan et al., 2011) and for anxiety (Weck et al., 2015). Furthermore, a study by Levy et al. (2016) provided preliminary evidence indicating that a strong therapeutic alliance can be developed with patients engaging in Virtual Reality Exposure Treatment (VRET) remotely. When studying the effect of the therapeutic alliance on treatment outcome, patients with a fear of flying associated therapeutic alliance positively with anxiety reduction (Meyerbroker & Emmelkamp, 2008). Moreover, in their study Wrzesien et al. (2012) found no difference in either of the three components of the therapeutic alliance between IVE and ARE treatment conditions. These results could be considered evidence of the impact of therapeutic alliance on different exposure treatments.

Another variable hypothesized to predict therapy outcomes is patients’ positive expectancy of treatment success (Constantino et al., 2018). Specifically in VRET, Price et al. (2008) found that positive expectancies of patients concerning the outcome may enhance the improvement of anxiety symptoms. In their study, subjects with a fear of flying and higher expectancies of treatment success presented stronger symptom reduction on self-report measures. Along the same line, a study conducted by Wu et al. (2020) with children undergoing exposure treatments, found that higher expectations were associated with higher youth mastery. Suggesting that if a child or adolescent expected positive outcomes, he/she persisted more time with the assignments and developed proficiency with repeated exposure tasks (Wu et al., 2020). Also, Bretón-López et al. (2015) found that participants in VRET had greater expectations regarding the success of treatment than those in a computer-aided exposure (CAE-SA) condition. This is evidence of expectations showing an important role in exposure-based treatments.

Considering that there has been no detailed investigation of the predictive power of the therapeutic alliance and treatment expectancies for exposure treatment success, the present study aims to close this gap.

The following hypotheses were addressed: (hypothesis #1) a better therapeutic alliance predicts better treatment outcome, (hypothesis #2) higher treatment expectations predict better treatment outcome, (hypothesis #3) a positive interaction between alliance and expectations enhances treatment outcome.

The present study resulted as a secondary outcome of an original RCT conducted by Botella et al. (2016). The main outcome paper showed no significant differences between the two treatment conditions in the long term. Taking those results into consideration we expect equal effects of the studied predictors on treatment outcome. However, analyses were run controlling for the effect of treatment conditions.

Methods

Participants

After advertisement, a total of 103 people contacted a Spanish University Clinic showing interest in taking part in the study. Out of these 103 people, 75 individuals were assessed for eligibility criteria in a first assessment session, after which twelve had to be excluded from the study because (1) they did not meet inclusion criteria (n = 2), (2) they refused to attend (n = 7), or (3) were not on call (n = 3). The total sample was N = 63, with 59 identifying as female (93.7%) and 4 as male (6.3%).

The data collection took place between January 2011 and January 2013. Inclusion criteria were: (a) meeting DSM-IV criteria for the diagnosis of SP (animal subtype) to cockroaches or spiders; (b) being at least 18 years old; (c) having a minimum 1-year duration of the phobia; (d) being willing to follow the study conditions and sign the consent form, and (e) presenting a score of at least 4 on the fear and avoidance scales of the diagnostic interview applied. Exclusion criteria were: (a) having another psychological problem that required immediate attention; (b) presenting current alcohol or drug dependence or abuse, psychosis or severe organic illness; (c) being currently treated in a similar treatment program; (d) being able to insert the hands in a plastic container with a cockroach or a spider (during the behavioral test); and (e) changing anxiolytics during the study (in the case of taking them).

The final sample had a mean age of 31.73 years (SD = 10.74), ranging from 20 to 70. About 49.2% of the sample was married and the other half (50.8%) was single. Most participants had a university degree (84.1%), and the others had finished high school (11.1%) or elementary school (4.8%). Two of the participants were taking anxiolytic drugs: one only occasionally due to a previous diagnosis of agoraphobia but not during the study period; the other to control binge-eating episodes in the framework of a previous eating disorder but kept the dose constant over the course of the study. Statistical analysis showed no differences between the two groups at pre-treatment on any demographic variables nor phobia duration or diagnostic variables. More information can be found in the main outcome paper (Botella et al., 2016).

Therapists

Five therapists were involved in the study, all of whom had a Ph.D. or a Master’s degree in Psychology. They were trained in CBT and had between 2 and 5 years of experience in the treatment of anxiety disorders and in the exposure technique (either IVE or ARE). In addition, they received training in the protocol from senior clinicians following the recommendations of Öst et al. (1991). Depending on the availability of therapists, the clinician who performed the initial and follow-up assessment was different from the one who conducted the treatment, while in other cases the clinician was the same. All therapists conducted both IVE and ARE treatments. They received weekly supervision from senior clinicians. Moreover, all assessment and treatment sessions were video recorded for supervision purposes.

Treatment

Participants received a one-session treatment guideline developed by Öst et al. (1991). The treatment included several therapeutic components applied in only one individual session lasting up to 3 h. First, patients received psychoeducation, then exposure hierarchy was elaborated with the patient and this exposure was started together with the questioning of the irrational beliefs that the patient may have. Treatment components consisted of exposure, modeling, reinforced practice, and cognitive challenge.

Treatment was applied in two different ways: in the IVE condition, participants were exposed to real cockroaches or spiders; in the ARE condition, participants were exposed to virtual animals (cockroaches or spiders) using the ARE system. As already mentioned, ARE is a variant of IVE that integrates virtual objects into the real world calculating the positions of the camera. Augmented Reality allows digital content using computer graphics mixed to overlay real-world objects. The mean treatment duration was 137 min (ranging from 62 to 180) in the IVE condition and 141.83 min (ranging from 70 to 180) in the ARE condition. As some exposure sessions took three hours, few participants in both conditions reported adverse effects such as: feeling tired, dizzy and having back pain. More information can be found in the main outcome paper (Botella et al., 2016).

Augmented Reality System and Hardware

In relation to the system, two devices were used to display Mixed Reality images: (1) AR 5DT HMD (head-mounted display) with an 800 × 600 resolution and a high (40°) fields of view where a USB Creative NX-Ultra camera is attached to the HMD to capture video stream; and (2) VR Goggles (Vuzix) that include two LCD devices with a 640 × 480 resolution and a 30° field of view and an embedded camera. The system includes 3D spiders and cockroaches and enables real-time interactivity. For more information see Botella et al. (2016).

Ethics Statement

Participants received information on informed consent during the admission session both verbally and in written form. Two consents were then provided by each participant. One regarding treatment and research participation and another including video-recorded authorization and data protection information.

Material

Patients went through a first session where they met the therapist. The therapist listened to their problem, and they agreed on treatment objectives, conditions, and participation. After this, therapeutic alliance and treatment expectations were assessed. Symptomatology was assessed through the BAT scale previously, after, and as a follow-up measure. All participants in both conditions received the same assessment. For the analysis, we considered fear, avoidance, and beliefs scores of the BAT dimensions. The assessment took place before and after treatment, as well as at 1-, 3- and 6-month follow-ups.

Behavioral Avoidance Test (BAT; Adapted by Öst et al., 1991)

The BAT is an observational measure used to assess the features of phobia in the context of exposure to the feared object, in order to obtain objective data about the person’s fear. Participants are then exposed to the phobic object in a stepwise manner (i.e., in 10 steps). Their performance on the test is scored by transforming the distance into a score rated on a scale. The assessment includes different dimensions: fear, avoidance, beliefs, maximum anxiety experienced by the participant, and performance and severity of the fear assessed by the therapist.

Working Alliance Inventory Short Version (WAI-S; Adapted by Corbella & Botella, 2003)

The WAI-S is made up of 4 subscales (a) agreement on tasks, (b) agreement on goals, (c) positive link, and (d) theory of change. The WAI-S consists of a total of 12 Items which the client scores according to the original 7-point Likert scale ranging from 1 (not at all) to 7 (totally). The total WAI-S score ranges from 12 to 84 points and showed good internal consistency (α Cronbach = 0.80).

Expectations and Satisfaction Regarding the Exposure Treatment (Adapted by Borkovec & Nau, 1972)

This questionnaire measures the participants’ expectations about the exposure component before the treatment and their satisfaction with it after the treatment. It includes six items rated from 0 (not at all) to 10 (very much). Items address how logical, satisfactory, recommendable, aversive and useful for the patient’s problems the treatment is. Some examples are: “To what extent do you think this program would be useful in your case?”, “To what extent do you find the treatment aversive?”, “To what extent are you satisfied with the program you are about to receive?”. It has shown good internal consistency (α Cronbach = 0.80). The adapted version of the scale has been used in previous studies (e.g., Baños et al., 2009; Botella et al., 2007, 2016).

Procedure

The study was based on a Randomized Controlled Trial (RCT), more detailed information can be found in the main outcome paper (Botella et al., 2016). First patients were assessed for eligibility criteria and were informed about the two treatment conditions and that they were going to be randomly assigned to one of the conditions. Then, patients were randomly assigned to receive either IVE (N = 31) or ARE (N = 32) through a computer-generated randomization list created by the “Random Allocation Software”; version 1.0. After this, patients had a first session where they gave written informed consent, and variables of interest were assessed (Therapeutic Alliance, Treatment Expectations, Symptomatology, etc.). Finally, they received a one-session exposure treatment, either IVE or ARE.

Statistical Analysis

Considering that in this study we have repeated measures nested within patients, we used multilevel models with a two-level structure (Level 1: repeated measures, Level 2: patients). Multilevel models account for the dependency of the data when there is a nested structure, they also have the advantage of handling missing data mimicking an intent-to-treat approach. This provides a more robust estimation of the parameters (Gómez Penedo et al., 2019; Raudenbush & Bryk, 2002). The analysis included the total scores of both assessed predictor variables (Therapeutic Alliance and Treatment Expectations). For each of the dependent variables (fear, avoidance, and beliefs scores) we ran the following models.

Time-As-Only-Predictor Models

To have an estimation of the rates of change during the follow-up period, we ran time-as-only-predictor models using BAT subscales as outcomes and including time (centered at the end of follow-up) as the only predictor. This model was conducted to estimate the change in outcome measures and for comparison purposes with the other models.

Level-1 Model

$$BAT_{ij} = \beta_{0j} + \beta_{1j} *\left( {{\text{Time}}_{ij} } \right) \, + r_{ij}$$

Level-2 Model

$$\begin{gathered} \beta_{0j} = \gamma_{00} + \gamma_{01} *\left( {{\text{PR}}\_{\text{BAT}}} \right) \, + u_{0j} \hfill \\ \beta_{1j} = \, \gamma_{10} + \, \gamma_{11} *\left( {{\text{PR}}\_{\text{BAT}}} \right) \hfill \\ \end{gathered}$$

At Level 1, BAT scores in subject j at time i, were predicted by the estimated scores of BAT at the end of follow-up for patient j (β0j) and the time effect during follow-up for patient j (β1j). At Level 2, β0j was predicted by the sample’s mean BAT score at the end of follow-up (γ00), while β1j was predicted by the sample average time effect during follow-up (γ10). We included baseline BAT to adjust for its effect both on the intercept and the slope.

Conditional Model to Determine Alliance and Expectations Main Effects

To estimate the expectation and alliance effects on BAT, we ran multilevel models with BAT scores (from post-treatment to the end of follow-up) as outcome. As level 1 predictor we incorporated time in months (centered at the end of follow-up). Then, we included expectations and alliance levels at baseline (grand-mean centered) as level 2 predictors of the intercept and the time effect. To control for initial levels of BAT we included baseline BAT as a level 2 predictor in the models. Furthermore, to adjust for condition effect we included condition coded as IVE = − .5; ARE = .5. We used the following equation:

Level-1 Model

$$BAT_{ij} = \beta_{0j} + \beta_{1j} *\left( {{\text{Time}}_{ij} } \right) \, + r_{ij}$$

Level-2 Model

$$\begin{gathered} \beta_{0j} =\, \gamma_{00} + \gamma_{01} *\left( {{\text{C}}\_{\text{Wai}}_{i} } \right) + \gamma_{02} *\left( {{\text{C}}\_{\text{Exp}}_{j} } \right) + \gamma_{03} *\left( {{\text{Condition}}_{i} } \right) \, + \gamma_{04} *\left( {{\text{PR}}\_{\text{BAT}}} \right) \, + u_{0j} \hfill \\ \beta_{1j} =\, \gamma_{10} + \gamma_{11} *\left( {{\text{C}}\_{\text{Wai}}_{j} } \right) + \gamma_{12} *\left( {{\text{C}}\_{\text{Exp}}_{j} } \right) \, + \gamma_{13} *\left( {{\text{Condition}}_{j} } \right) \, + \gamma_{14} *\left( {{\text{PR}}\_{\text{BAT}}} \right) \hfill \\ \end{gathered}$$

At Level 2, as targeted independent variables we included baseline alliance scores (grand-mean centered; C_Wai) as a predictor of the intercept (γ01) and the time slope (γ11), and baseline expectations (grand-mean centered; C_Exp) as a predictor of the intercept (γ02) and time slope (γ12). To adjust for treatment effects, we included the condition as a further predictor of the intercept (γ03) and the slope (γ13). Furthermore, we included baseline BAT as a predictor of both the intercept (γ04) and slope (γ14).

Interactive Effects of Expectations and Alliance with Treatment

Finally, we tested the interactive effect of expectations by therapeutic alliance on BAT, adjusting for treatment condition and previous BAT scores. For this model, we used the following equation:

Level-1 Model

$$BAT_{ij} = \beta_{0j} + \beta_{1j} *\left( {{\text{Time}}_{ij} } \right) \, + r_{ij}$$

Level-2 Model

$$\begin{gathered} \beta_{0j} =\, \gamma_{00} + \gamma_{01} *\left( {{\text{C}}\_{\text{Wai}}_{j} } \right) \, + \gamma_{02} *\left( {{\text{C}}\_{\text{Exp}}_{j} } \right) \, + \gamma_{03} *\left( {{\text{C}}\_{\text{Exp}}\_{\text{X}}\_{\text{C}}\_{\text{Wai}}_{j} } \right) \, + \gamma_{04} *\left( {{\text{Condition}}_{j} } \right) \, + \gamma_{05} *\left( {{\text{PR}}\_{\text{BAT}}} \right) \, + u_{0j} \hfill \\ \beta_{1j} =\, \gamma_{10} + \gamma_{11} *\left( {{\text{C}}\_{\text{Wai}}_{j} } \right) \, + \gamma_{12} *\left( {{\text{C}}\_{\text{Exp}}_{j} } \right) \, + \gamma_{13} *\left( {{\text{C}}\_{\text{Exp}}\_{\text{X}}\_{\text{C}}\_{\text{Wai}}_{j} } \right) \, + \gamma_{14} *\left( {{\text{Condition}}_{j} } \right) \, + \gamma_{15} *\left( {{\text{PR}}\_{\text{BAT}}} \right) \hfill \\ \end{gathered}$$

At Level 1, BAT scores at time i for patient j were predicted by patient j’s estimated BAT score at the end of follow-up (β0j) and the time effect during follow-up for patient j (β1j). At Level 2, as targeted independent variables we included baseline alliance scores (grand-mean centered; C_Wai) as a predictor of the intercept (γ01) and time slope (γ11), baseline expectations (grand-mean centered; C_Exp) as a predictor of the intercept (γ02) and the time slope (γ12), and the interactive effect of expectations by alliance both on the intercept (γ03) and the time slope (γ13). As in the conditional models, to adjust for treatment effects, we included condition as a further predictor of the intercept (γ04) and the slope (γ14). Furthermore, we included baseline BAT as a predictor of both the intercept (γ05) and slope (γ15).

Results

Sample Descriptives

Descriptive statistics of treatment expectations, therapeutic alliance and BAT for both conditions separately are presented for baseline, post-treatment and follow-up period in Table 1.

Table 1 Sample descriptive for assesment at baseline and last follow-up
Full size table

Time-As-Only-Predictor Model

Results showed that participants had an estimated level of fear of 4.31 at the end of follow up, γ00 = 4.31, SE = 1.32, 95% CI [1.71, 6.91], t(69) = 3.27, p < .001. The model also showed an approach significance monthly reductions in the fear subscale, γ10 = − 0.20, SE = 0.11, 95% CI [− 0.42, 0.01], t(189) = − 1.88, p = .06. Every month there was a 0.20 units reduction in fear over the course of follow up.

In relation to the avoidance subscale, participants had an estimated level of avoidance of 2.82 at the end of the follow up, γ00 = 2.82, SE = 1.79, 95% CI [− 0.71, 6.34], t(67) = 1.57, p = .12. The model also showed an approach significance monthly reduction of 0.22 units in the avoidance subscale, γ10 = − 0.22, SE = 0.19, 95% CI [− 0.45, 0.01], t(189) = − 1.85, p = .07.

Finally, participants had an estimated level of beliefs of 2.34 at the end of follow up, γ00 = 2.34, SE = 1.12, 95% CI [0.14–4.55], t(71) = 2.1, p < .005. The model also showed an approach significance monthly reduction of 0.19 units in the belief subscale, γ10 = − 0.19, SE = 1.11, 95% CI [− 0.40, 0.02], t(189) = − 1.74, p = .08.

Conditional Model

In these models, we analyzed alliance and expectation effects on the three targeted dimensions of the BAT (i.e., fear, avoidance, and beliefs). Therapeutic alliance was first analyzed with WAI-S sub-scales, as results did not vary and for not adding more multiple comparisons, we decided to report the final models with the WAI-S total scores. Full results of the models are presented in Table 2 for the fear subscale, Table 3 for the avoidance subscale, and Table 4 for the beliefs subscale.

Table 2 Results of the unconditional model, expectations and therapeutic alliance main effect model and interactive effects model of expectations and therapeutic alliance by follow-ups for fear subscale
Full size table
Table 3 Results of the unconditional model, expectations and therapeutic alliance main effect model and interactive effects model of expectations and therapeutic alliance by follow-ups for avoidance subscale
Full size table
Table 4 Results of the unconditional model, expectations and therapeutic alliance main effect model and interactive effects model of expectations and therapeutic alliance by follow-ups for beliefs subscale
Full size table

In the models with the fear subscale, we didn’t find a significant effect of the alliance on the last follow up score (γ01 = − 0.03, SE = 0.73, 95% CI [− 1.46, 1.40], t(141) = − 0.04, p = .97) nor of expectations (γ02 = − 0.20, SE = 0.35, 95% CI [− 0.89, 0.49], t(133) = − 0.56, p = .57). Furthermore, we neither found a significant effect of alliance (γ11 = − 0.25, SE = 0.19, 95% CI [− 0.62, 0.12], t(168) = − 1.35, p = .18) nor expectations (γ12 = − 0, SE = 0.09, 95% CI [− 0.18, 0.17], t(168) = − 0.04, p = .97) on the slope of the fear subscale.

In the models with the avoidance subscale, we did not find a significant effect of the alliance (γ01 = 0.02, SE = 0.84, 95% CI [− 1.64, 1.69], t(134) = 0.03, p = .98) but we did find an approach significant effect of expectations on the last follow up score (γ02 = − 0.74, SE = 0.40, 95% CI [− 1.52, 0.04], t(13) = − 1.88, p = .06). Furthermore we did neither find a significant effect of alliance (γ11 = − 0.25, SE = 0.19, 95% CI [− 0.62, 0.12], t(168) = − 1.35, p = .18) nor expectations (γ12 = − 0, SE = 0.09, 95% CI [− 0.18, 0.17], t(168) = − 0.04, p = .98) on the slope of the avoidance subscale.

In the models with the belief subscales, we did not find a significant effect of the alliance, γ01 = 0.42, SE = 0.70, 95% CI [− 0.95, 1.80], t(153) = 0.61, p = .55 but we did find a significant effect of expectations on the last follow up score, γ02 = − 0.94, SE = 0.33, 95% CI [− 1.45. − 0.35], t(153) = − 2.88, p < .01 on this specific dimension. Moreover, we did neither find a significant effect of alliance (γ11 = 0.06, SE = 0.19, 95% CI [− 0.31, 0.43], t(168) = 0.33, p = .74) nor expectations (γ12 = − 0.05, SE = 0.09, 95% CI [− 0.23, 0.12], t(168) = − 0.60, p = .55) on the slope of the belief subscale.

Interactional Models

Interactive Effect Model of Therapeutic Alliance and Treatment Expectation

Interactional models for fear did not show a significant interactive effect of alliance by expectations neither on the fear subscales’ last follow-up score (γ03 = 0.02, SE = 0.43, 95% CI [− 0.82, 0.87], t(140) = 0.05, p = .99) nor on the slope (γ13 = 0.20, SE = 0.11, 95% CI [− 0.02, 0.41], t(168) = 1.76, p = .08).

Interactional models for avoidance did not show significant interactive effects of alliance and expectations on the avoidance subscales’ last follow up score (γ03 = 0.36, SE = 0.50, 95% CI − 0.64, 1.35], t(131) = 0.71, p = .48) but it did on the slope (γ13 = 0.25, SE = 0.11, 95% CI [− 0.02, 0.41], t(168) = 1.98, p < .05).

Interactional models for beliefs did not show significant interactive effects on the belief subscales’ last follow up (γ03 = 0.57, SE = 0.41, 95% CI [− 0.25, 1.38], t(152) = 1.38, p = .17) but it did on the slope (γ13 = 0.23, SE = 0.11, 95% CI [0.01, 0.45], t(168) = 2.07, p < .05).

Discussion

The goal of the present study was to investigate the effects of the therapeutic alliance and treatment expectations on exposure treatment outcomes. We analyzed (1) the effects of the therapeutic alliance on treatment outcome, (2) the effects of treatment expectations on treatment outcome, and (3) the interaction between both variables on treatment outcome.

Contrary to hypothesis one, results indicate that the therapeutic alliance is not a direct predictor of treatment outcome for SP. In accordance with hypothesis two, treatment expectations meaningfully predicted beliefs and avoidance symptoms related to SP. With regard to hypothesis three, there was no interaction effect of treatment expectations and the therapeutic alliance on symptom scores over follow-up.

With regard to the therapeutic alliance, our results suggest that it is not a significant predictor of symptom change in a one-session exposure treatment for SP. Our findings are in line with other studies on anxiety disorders that measured the therapeutic alliance and found no significant alliance-outcome relationship (e.g., Maiwald et al., 2019). It might be that the alliance is not equally important in the treatment of this diagnostic group as it is for other mental disorders. Our results, however, are not in line with the findings by Buchholz and Abramowitz (2020) and Pan et al. (2011) who presented the alliance as a possible estimator of treatment outcome in a one-session exposure for SP. Specifically, the authors found that the therapeutic alliance had an effect on avoidance symptoms (Pan et al., 2011). Moreover, as the quality of the therapeutic alliance can vary over time, differences in results may depend on the moment of assessment. In the study conducted by Pan et al. (2011), the alliance was assessed immediately following treatment, and by a blind observer after viewing a videotape of the therapy session. This is an important distinction considering that in the present study the therapeutic alliance was completed by the patient in the first session prior to the intervention. The non-significant alliance outcome association in our findings may be related to this. Moreover, different variables such as adherence to treatment, self-efficacy, cognitive change, and motivation can mediate the relationship between the therapeutic alliance and treatment outcomes which can lead to a discrepancy (Böhnlein et al., 2019; Buchholz & Abramowitz, 2020). Another issue that may have influenced the non-significant alliance results is the fact that treatments were provided by 5 different therapists. Results may benefit from including the therapist’s effect on the analysis regarding evaluations of therapeutic alliance, which wasn’t included in this study.

With regard to treatment expectations, positive expectancies of patients enhanced the improvement of symptoms which is in line with Price et al. (2008) findings. Moreover, results suggest that expectancies are equally important in IVE as in ARE. Effects of expectancies on beliefs and avoidance symptoms are likely to be related to the fact that confidence in the treatment may produce a change in cognition. Considering cognitive-behavioral theory, changes in beliefs would help the patient to feel more capable, comfortable, and with a greater sense of personal competence when coping with fear versus avoiding it (Beck et al., 1979). Moreover, these results correspond with the idea of transforming cognitions to help the patient confront the anxiety activated by the phobic object.

Finally, the lack of interaction between expectations and the therapeutic alliance is likely related to the fact that the alliance alone was not a significant predictor of change.

There are several limitations to bear in mind when interpreting our results. Since this was a study with a one-session treatment, results in other types of treatments could vary. Also, therapeutic alliance and expectations were measured at baseline as predictors of outcomes. This is important considering that the therapeutic alliance was assessed only prior to the intervention and in that sense differs from previous studies. Including an assessment of the therapeutic alliance after the intervention could provide useful additional information, as several assessments of the predictors over time may lead to divergent results. Finally, because of the size of the sample, the focus on a specific subtype of SP, and the previously mentioned limitations such as lack of information on previous treatments, the generalizability of our results is limited.

Further research is necessary to address the outlined limitations. More process-outcome studies for patients with SP will help to understand the relational change mechanisms involved in different exposure treatments. Also, we need to consider that technology is continuously developing and evolving, graphics and designs in AR improves which may add value to the exposure devices and as a consequence influence new treatments. Also, the present research was based on one session of exposure treatment that turned out to be long. We recommend taking breaks during the session if it is prolonged.

Overall, the results of this study suggest that expectations have a meaningful effect on outcomes in exposure-based treatments, both in IVE and ARE. This underlines the importance for therapists to consider and foster positive expectations regardless of the type of exposure they work with.