A comparison of virtual reality and verbal imaginal exposure for childhood anxiety disorders

Abstract

Virtual reality is a promising tool for improving efficacy of exposure therapy for childhood anxiety disorders (CADs), particularly for exposures that are challenging to orchestrate in vivo. The present study compared virtual reality exposure (VRE) to verbal imaginal exposure (IE) on anxiety elicited, homework completion, and preparation for subsequent exposures. Forty-five youth with CADs completed both types of exposure in session (order randomly assigned), practiced one form of exposure as homework (randomly assigned), and returned a week later to repeat both exposures (in randomized order), provide user feedback, and complete a related in vivo exposure. Both VRE and IE elicited moderate anxiety that decreased to mild during the exposure and were associated with lower initial anxiety ratings in subsequent exposures. These patterns did not differ across exposure type. The two techniques did not differ on homework completion. VRE was associated with greater simulator (motion) sickness symptoms. Participants rated VRE as more realistic than IE and noted pros and cons for both approaches. The present study supported use of VRE and IE in treatment of CADs without demonstrating superiority of either technique in effectiveness or homework completion.

A substantial body of research and practice supports the use of cognitive-behavioral therapy (CBT) for the treatment of childhood anxiety disorders (CADs) and obsessive-compulsive disorder (OCD; here forth jointly referred to as CADs (Banneyer et al., 2018; Ewing et al., 2015; Kreuze et al., 2018; Skarphedinsson et al., 2015; Wang et al., 2017). With growing evidence supporting exposure as the active ingredient (Ale et al., 2015; Bilek et al., 2022; Peris et al., 2017; Whiteside et al., 2020), efforts to develop tools that aid or expand the use of exposure hold promise for improving the efficacy, effectiveness, and accessibility of CBT for CADs. One such avenue is using virtual reality (VR), which can deliver exposure to stimuli that cannot be recreated on demand in the office or in standard live settings. As such, the current study examines the potential for virtual reality exposure therapy (VRET) to improve CBT for CADs.

Exposure therapy involves prolonged contact with feared stimuli to allow the patient to learn that the stimuli are not as dangerous and are more manageable than previously believed. Despite the support for the importance of exposure, there is not a consensus on the underlying mechanism of action with the most prominent theories including habituation, emotional process, and inhibitory learning (Whiteside et al., 2020). Uniting these theories are the agreement that, to be therapeutic, exposures need to evoke anxiety and be continued long enough for new learning to take place, whether that is defined as reduction of anxiety ratings or changes in fear inducing expectations. Exposures that are more engaging and realistic (e.g., actively completing an activity) are generally thought to be more effective than those that are more passive or contrived (e.g., imaging a situation or role playing; Gillihan et al., 2012).

One challenge to the successful implementation of exposure therapy is the inability to replicate some feared stimuli in vivo. Some feared stimuli are physically impossible to create, such as combat or natural disasters for treatment of PTSD, or are highly impractical, such as flying in a plane for phobias. Other stimuli occur commonly, but are difficult to initiate on demand, such as encountering a large spider, being in a crowded school hallway, or speaking with or in front of classmates. In some circumstances these fears can be most appropriately and directly addressed through thought or imaginal exposures (IE), such as recounting a traumatic memory for PTSD or repeating intrusive thoughts for OCD. Other times, situations can be sufficiently approximated through role plays with confederates or, if available, practice with other patients in group treatment settings, such as giving presentations or starting conversations. Less ideally, sometimes the only available options are IE with or without the aid of visual prompts (e.g., videos) with the intention of increasing a patient’s confidence to complete in vivo exposures when the moment arises.

VRE presents the potential to deliver highly immersive experiences to stimuli that cannot be replicated on demand in vivo. The theoretical underpinning of VRE is that this immersive quality approximates real life experience to a greater degree than alternative approaches, such as role playing or verbal IE. Based on research that has been largely conducted with adults, VREs have been successfully developed and tested for combat, flying, and large spiders (Maples-Keller et al., 2017). More recently, VRE has been applied to more ordinary, but still difficult to design, exposures such as presentations, social interactions, and procedures with needles (Maples-Keller et al., 2017). The nascent research applying VR to mental health treatment for children and adolescents has thus far shown potential for its use in exposure therapy for anxiety disorders including social phobia and specific phobias (Blanco et al., 2023). A major limitation of this work has been the paucity of studies utilizing randomly assigned comparator conditions, particularly comparison to alternative active intervention (Blanco et al., 2023). One exception is a previous study in which we explored the possibility of using a classroom scenario to address fears of academic failure in GAD in which VRE was compared to verbal IE (Whiteside et al., 2020). In this study, although both VRE and verbal IE elicited anxiety during the exposures, they did not differ in the amount of anxiety, failing to support the hypothesis that VR would produce more evocative exposures (Whiteside et al., 2020). However, the conclusions of this study are limited by the narrow clinical content, the possibility that IE is better suited for worry than other types of anxiety, and limited immersiveness of VR technology at the time of the study, namely, participants found the use of animation limited the realism of VRE in the earlier study.

Since the time of our initial study comparing VRE and verbal IE, technological advancements have simplified the creation and deployment of realistic VR content. Modern 360-degree cameras now capture realistic VR video with minimal post-processing required. Stand-alone VR headsets now have powerful processors to play 360-degree VR videos while also detecting and reacting to spoken verbal cues and head pose/gaze-directed input from the user (Zainal et al., 2021). These immersive visual and auditory experiences and opportunities for sound- and motion-activated interaction increase the potential for VRE to have an advantage over verbal IE while having similar portability and accessibility as verbal IE. One potential downside of the new immersive capabilities is the possibility for them to trigger simulation sickness symptoms (Kennedy et al., 1993; Shahid et al., 2024). However, extant work with this technology seems to indicate reduced or equivalent sickness symptoms compared to earlier versions of VR (Zainal et al., 2021).

Given the limitations of the earlier study, the current experiment tested the potential to improve CBT for CADs through VRE with realistic live action videos covering a variety of exposure situations that are difficult to implement in the office. To address limitations from the lack of active comparator groups in research thus far using VRE with children, the efficacy of VRE was compared to the most likely alternative, verbal IE. Specifically, we compared the two types of exposure in effectiveness for eliciting anxiety during in-session exposures, facilitating homework completion, and preparation for subsequent exposures. Consistent with the principles underlying successful exposure therapy, the current study was designed to test the following hypotheses: 1) Exposures with both VRE and IE would successfully elicit anxiety that decline during the exposure, but VRE would elicit higher levels of anxiety than IE; 2) VRE would engage patients in more home (between-session) exposures than would IE; and 3) Home exposure practice with VRE would lead to lower anxiety ratings during subsequent exposures, including an in vivo exposure, relative to home practice with IE. Given prior reports of simulation sickness symptoms with VR, we assessed simulation sickness symptoms following each exposure to explore if symptoms experienced were unique to VRE or present for other reasons (e.g., nervousness). Finally, given calls for greater attention to user experience when testing effectiveness of VRE (Shahid et al., 2024), we interviewed children and their parents about their experiences with VRE and verbal IE.

1. Methods

1.1. Participants

Participants consisted of 45 children (identified gender: 37 girls, 82 %; 7 boys, 16 %; 1 transgender youth, 2 %) age 7–17 (M = 13.24 ± 2.73) and their parents (41 mothers, 4 fathers) participating in exposure therapy for treatment of CADs in the [institution] Pediatric Anxiety Disorders Clinic (PADC). Demographics and baseline anxiety symptom scores are summarized for the study sample and by condition in Table 1. The most common primary diagnoses were social phobia (n = 21), generalized anxiety disorder (n = 9), obsessive compulsive disorder (n = 5) and separation anxiety disorder (n = 5). Comorbidity was common (M_diagnoses = 2.38 ± 1.30, range 1–6). Of the 45 child-parent pairs who enrolled and completed the first study session, 40 completed all study sessions. Reasons for non-completion included other commitments and not completing the homework for the study. One participant exited the study due to feeling overwhelmed during Session 1 after writing the IE story and before reading it aloud for the exposure.

Table 1.

Study participant demographics by homework condition.

Demographic	Category	Total Sample	IEa	VREa
Sample Size		45	20	25
Gender at Birth	Female	38 (84.4 %)	17 (85.0 %)	21 (84.0 %)
	Male	7 (15.6 %)	3 (15.0 %)	4 (16.0 %)
Identified Gender	Female	35 (77.8 %)	16 (80.0 %)	19 (76.0 %)
	Male	8 (17.8 %)	3 (15.0)	5 (20.0 %)
	Did not disclose	1 (2.2 %)	1 (5.0 %)	-
	Other	1 (2.2 %)	-	1 (4.0)
Race/Ethnicity	Black	2 (4.4 %)	1 (5.0 %)	1 (4.0 %)
	Asian	2 (4.4 %)	1 (5.0 %)	1 (4.0 %)
	White	41 (91.1 %)	-	1 (4.0 %)
	Hispanic/Latinx	1 (2.2 %)	18 (90.0 %)	22 (88.0 %)
	Mean (SD)	Median (Range)	Mean (SD)	Mean (SD)
Age at consent	13.2 (2.8)	13 (7, 17)	13.1 (3.0)	13.3 (2.6)
SCAS—Child	47.9 (21.1)	45 (8, 86)	48.5 (20.8)	47.6 (21.9)
SCAS—Parent	33.1 (15.6)	31 (5, 72)	34.3 (16.5)	32.3 (15.2)

Note.

^aIE (imaginal exposure) and VRE (virtual reality exposure) refers to the homework condition to which the participants were assigned.

SCAS = Spence Children’s Anxiety Scale, Parent- and Child- Report.

1.2. Procedure

The [institution]’s Institutional Review Board approved the research protocol and materials. PADC staff invited children and their caregivers to participate after their initial assessment in the PADC if the child had a qualifying diagnosis (i.e., separation anxiety disorder, social phobia, generalized anxiety disorder, panic disorder with or without agoraphobia, specific phobia, or obsessive-compulsive disorder) and the child endorsed fears related to at least one VRET content area on the study screening questionnaire. All patients in the PADC complete a standardized diagnostic assessment that includes the Mini-International Neuropsychiatric Interview for Children and Adolescents (MINI-KID; Sheehan et al., 2010) and Spence Child Anxiety Scales, Parent & Child Report (Muris et al., 2002, 2000; Spence, 1998). They then begin treatment with psychoeducation on parent-coached exposure therapy for anxiety and OCD, create a treatment plan (i.e., fear hierarchy), and proceed to completing parent-coached exposures, typically in the second treatment session. Families who consented to study participation began the study after completing at least one treatment session with parent-coached exposure to ensure they were familiar with the exposure process. One family completed two sessions with parent-coached exposure prior to the first study session; the other participating families completed three or more exposure sessions prior to study participation.

Study staff provided written descriptions of the study and spoke with youth and parents in person or by telephone to explain the study and obtain informed assent and consent. Prior to the first study visit for each participant, two random assignment procedures determined (1) the order of in-session exposure presentation (i.e., VRE then IE or IE then VRE) and (2) the home exposure practice condition (VRE or IE). Participants attended two study sessions, each approximately 60-min and spaced a week apart. In Session 1, study staff described the three related exposures assigned to the participant for the study based on the content area the child indicted as most anxiety provoking on the study screening questionnaire (i.e., a VR video, analogous IE, and analogous in vivo exposure). The study staff then asked the child to rate how nervous they would be to do each exposure using a scale of 0 (not nervous at all) to 10 (extremely nervous). Table 2 lists exposure content. In Session 1, children completed the VRE and verbal IE, with order of presentation determined by random assignment. The VRE was presented with a a Pico Goblin headset. Parents viewed the VRE video simultaneously on a tablet. For the IE, children were instructed to write a story of 3–5 sentences about their fear coming true and then read it aloud repeatedly.

Table 2.

Assignment of virtual reality (VR), imaginal, and in vivo exposures by fear category.

Feared Situation	VR Exposure	Imaginal Exposure	In Vivo Exposure	n for VR Homework	n for IE Homework
Being observed in school hallway (social)	Stand in busy hallway at school	Being seen or watched by others your age at school	Drop a full handful of coins in the clinic lobby repeatedly	0	3
Answering a question in class (social)	Answer persuasive writing project question in classrooma, b	Answering a question in class	Answer question in front of others (3—4), without knowing the question beforehand.	0	1
Class presentation (social)	Give class presentation in front of classrooma	Giving a class presentation	Give a 2–3 min presentation to clinic staff in secretary’s office (3–4 people)	7	4
Doing poorly on an assignment (performance worry)	Receive assignment with low grade from teacher in classrooma	Getting a low grade on a school assignment	Role play: complete math worksheet and an adult corrects it	2	2
Being around peers in a classroom (social)	Sit at desk in classroom with students at desks taking testsa	Being around others your age in a classroom	Sit in the pediatrics waiting room alone	0	0
Asking a teacher a question in front of the class (social)	Asking the teacher to go to bathrooma, b	Asking a teacher a question in front of your class	Go into secretary’s office and ask where the bathroom is	1	0
Being observed by peers in the school lunchroom (social)	Standing in school lunchroom	Being seen or watched by others in the lunch room	Eat in the patient cafeteria	2	0
Being in a busy cafeteria (agoraphobia)	Standing in school lunchroom	Being in a busy cafeteria	Eat in the patient cafeteria	1	1
Conversations with peers (social)	Conversation with peers at lunchroom tableb	Having a conversation with others your age	Start a conversation with similar aged person in the pediatrics lobby	3	2
Shots (needle phobia)	Clinician gives patient an injection	Getting a shot	Role play getting a shot	2	2
Blood draws (needle phobia)	Clinician draws blood from a patient	Getting a blood draw	Role play getting a blood draw	2	0
Busy sidewalks (agoraphobia)	Busy city sidewalk scene	Feeling afraid in a busy public space	Walk alone in crowded hallway in the clinic lobby	4	1
Getting lost in public (separation)	Busy city sidewalk scene	Getting lost in public	Walk alone in crowded hallway in the clinic lobby	0	1

Note. Some VR videos could be used for multiple fears; in these cases, the imaginal and in vivo exposures related to the top fear the participant indicated on the study screening questionnaire.

^aTwo versions of the same video were available, one with child actors predominant in the scene and the other with adolescent actors predominant.

^bVideo featured an audio-activated response such that the teacher/peer spoke after an audible verbalization from the participant.

Following common practice in exposure therapy, youth provided subjective ratings of distress (SUDS, 0–10) at the onset of exposures and then periodically (approximately every 60–90 s). For VR exposures, youth entered a SUDS rating using the VR headset and verbally told study staff the rating. For IE and the in vivo exposures, study staff prompted and recorded SUDS ratings. Exposures continued until discontinuation rules were met: a) anxiety decreased by 50 % from initial anxiety for two consecutive trials, b) ratings of 0 for two consecutive trials, or c) 10 trials. Children completed the Simulator Sickness Questionnaire following each exposure. At the conclusion of Session 1, study staff told the family which homework condition (VR or IE) had been randomly assigned, provided the corresponding exposure stimulus (i.e., headset with assigned exposure or the participant’s verbal IE story), encouraged the family to complete the assigned exposure daily, and provided verbal and written instructions for how to use the exposure tracking form to record exposures during home practice.

In Session 2, families returned completed exposure homework equipment and tracking forms, and children completed the same two exposures as Session 1 presented in randomized order and the Simulation Sickness Questionnaire after each exposure. Children and parents reported on their experience via semi-structured interview with study staff. Following the interview, study staff led the child through the assigned in vivo exposure. Each child participant was remunerated $50 for study completion.

1.3. Materials

1.3.1. VR equipment

Youth completed VR exposures using a stand-alone VR headset called the Pico Goblin which has a resolution of 2560 × 1440, refresh rate of 70 Hz, adjustable IPD, and 92-degree FOV. The headset ran a custom app developed by Limbix VR, which was later sold to BehaVR and is now part of RealizedCare. Parents were given a small Bluetooth-connected tablet that showed a real-time, low-resolution view of what the youth could see in the headset. Participants stored, transported, and charged the VR headset and tablet in a rugged case designed by Minnesota HealthSolutions.

1.3.2. Exposure content

The authors of this paper developed a set of VR exposures that would be applicable to a broad subset of child and adolescent fears with a focus on stimuli that are challenging to replicate in the office. To do so, clinic therapists brainstormed exposure content families often say they wish they could create during sessions, including social interactions with peers and teachers especially in school settings; busier public spaces for separation anxiety, agoraphobia, and general safety worries; and receiving a bad grade in a school setting, a common worry for youth with GAD in the clinic. Then, creation of more frequently needed content was prioritized over less commonly needed content. The final list of exposure content is summarized in Table 2. Variations for younger and older youth were filmed using different aged actors. Some VR exposures were interactive, requiring youth to give verbal input before progressing to the next section (answering a question in class, giving a 3-minute speech, asking to use the bathroom, interacting with peers in the lunchroom). Other VR exposures were passive, allowing youth to experience a situation (getting a bad grade on an assignment, taking a test in class, standing in the school lunchroom or hallway, standing on a street corner, watching a vaccination, watching a blood draw). Limbix VR completed final script authoring and video production, ensuring that the VR videos were compatible with the VR headset.

1.3.3. Clinical diagnoses

During the initial clinical assessment, PADC clinicians used the Mini International Neuropsychiatric Interview for Children and Adolescents (MINI-KID; Sheehan et al., 2010) to establish the presence of an anxiety disorder. The MINI-KID is a brief structured diagnostic interview for psychiatric diagnoses in children and adolescents aged 6–17 with substantial to excellent concordance with other structured interviews. Clinicians implemented only the modules for assessing the anxiety disorders (including OCD and post-traumatic stress disorder), major depression, attention deficit hyperactivity disorder, and oppositional defiant disorder.

1.3.4. Anxiety symptoms

Parents and youth completed the Spence Children’s Anxiety Scale-Child and Parent Versions (SCAS-C/P) to assess severity of the youth’s anxiety and OCD symptoms. The SCAS-C/P are Likert-type questionnaires with robust reliability and validity measuring six facets of anxiety in children and adolescents. The current study used the total score as a measurement of overall anxiety disorder symptoms (Muris, Schmidt, & Merckelbach, 2000; Muris, Merckelbach, Ollendick, King, & Bogie, 2002; Spence, 1998).

1.3.5. Anxiety ratings

Youth rated their anxiety during exposures using a 0–10 scale consistent with the commonly used clinical tool, subjective units of distress (SUDS). Study staff presented a sheet of paper with three components: a thermometer graphic with numbers 0 through 10; qualitative descriptors (relaxed no anxiety, a little anxiety, some anxiety, a lot of anxiety, most anxiety ever); and a series of five facial expressions indicating increasing levels of anxiety.

1.3.6. Simulator sickness symptoms

The Simulator Sickness Questionnaire (SSQ-m) is a 16-item questionnaire assessing physical symptoms similar to motion sickness that can arise when using simulators of a visual environment such as driving and flying simulators and VR systems (Kennedy et al., 1993). The SSQ is the most common tool to assess simulator sickness symptoms and is well-validated, including demonstrated ability to predict drop-out from simulator studies (Balk et al., 2013). For the current study, simple language synonyms were added in parentheses following items with higher-level reading words (e.g., “sick to stomach” after “nausea”). Each item was scored 0–3, with higher numbers indicating greater symptom severity. For this study, a total SSQ-m score was calculated as the sum of all items (Bouchard et al., 2009).

1.3.7. User experience interviews

Study staff interviewed youth and parent participants together about their experience with the VR and IE exposures following a semi-structured interview guide. Interviews lasted approximately 20 min. The open-ended questions in the interview inquired about the child’s and parent’s opinions about VR and IE exposures in terms of effectiveness in eliciting anxiety similar to real life, potential for aiding or interfering with exposure therapy, and recommendations for improvement. The interview guide included prompts to elicit responses from both children and parents. The interview also elicited three quantitative ratings on a 0–10 rating scale, with 5 being neutral: which type of exposure made the youth more nervous, writing (0) or VR (10); degree to which the youth enjoyed exposure homework compared to how they usually feel about it, enjoyed less (0) to enjoyed more (10); and the parent’s effort to get their child to complete exposure homework compared to usual, much harder (0) to much easier (10). Interviews were audio recorded and transcribed verbatim. The full interview guide is available from the authors upon request.

1.4. Planned analyses

1.4.1. Statistical analyses

Group differences by homework condition on demographic variables and pre-clinical assessments were tested using Chi-square tests, Fisher’s Exact test, and two-sample t-tests. For Hypothesis #1, longitudinal models using both last-observation-carried-forward imputation (LOCF) and using all available data tested change in SUDS ratings over the course of each exposure (pre-rating and 1–10). These models examined exposure type and homework condition and potential interactions between exposure type and HW condition, adjusted for rating number (pre, 1–10). Further, paired t-tests tested for differences between the two exposure types within each study session in SUDS at various time points (initial, maximum, and final), change in SUDS (initial to final), and duration of exposures (in minutes and number of ratings). For Hypothesis #2, two-sample t-tests examined group differences by homework condition in exposure homework completion. For Hypothesis #3, the differential ability of VRE and IE to prepare patients for subsequent exposures was examined in two ways. First, order effects were explored based on which exposure type was presented first within each study session. Specifically, SUDS ratings (initial, max, final), change in SUDS ratings, and exposure duration (in minutes and number of ratings) for the second exposure completed in each study session were compared across groups designated by the first-presented exposure type (VRE vs IE) using two-sample t-tests and longitudinal (repeated measures) mixed models adjusted for rating number. Second, two-sample t-tests compared homework conditions (VRE vs IE) on max SUDS ratings during the Session 2 crossover exposure (i.e., IE for participants in the VRE homework condition and VRE for participants in the IE homework condition) and during the generalization in vivo exposure. An additional multivariable linear regression model with a covariate adjusting for Session 1 maximum SUDs tested homework condition differences on the cross-over and in vivo max SUDS ratings, accounting for any potential first session differences or effects. To use all available data during the session, longitudinal models with the outcome of either VRE or IE exposure SUDS were used to test for an effect of the exposure type order (VRE or IE first), adjusted for homework condition, and rating number. Pairwise interaction terms between order, rating number, and exposure type order were tested in these models, which were run for both all available SUDS ratings, and LOCF imputation models. To test whether VRE resulted in more simulation sickness symptoms, a longitudinal (repeated measures) mixed model examined differences in SSQ-m scores (simulator sickness symptoms) between the exposure types, order (comparing SSQ-m score from first exposure completed with scores from subsequent exposures, i.e., 2nd, 3rd, 4th), and homework condition (VRE vs IE). Regarding user experience ratings, a single sample t-test examined whether participant ratings for which exposure was more realistic differed significantly from neutral (rating of 5 on scale from 1 to 10 with 1 = IE most realistic to 10 = VRE most realistic). Quantitative analyses were conducted using SAS (version 9.4; Cary, NC) and plots were created using the ggplot2 package (Wickham H, 2016) in Rstudio (version 4.2.2).

1.4.2. User experience thematic analysis

Thematic analysis of children and parents’ interview responses was used to gain a nuanced understanding of the experiences of participants regarding VRE and IE. Our approach to thematic analysis was guided by the recommendations of Braun and Clarke (2006). Two investigators not involved in delivering the intervention independently read transcripts and collaboratively developed a coding plan integrating deductive and inductive codes. Based on our team’s prior study with VR and other VR user experience research (Shahid et al., 2024), we used deductive codes relating to acceptability, feasibility, technological challenges, technological novelty, and anxiety intensity, including the degree to which exposure content felt real. Inductive codes, created based on transcript content, were importance of the unexpected/need for varied content (VR), importance/lack of realism (VR), new approach to exposures (VR), comparison to a game (VR), boring (VR), freedom from real-life consequences (VR), comparison to therapy (IE & VR), and accountability/adherence due to tracking forms/research (IE & VR). Transcripts were coded independently (codes compared for agreement), and themes were identified collaboratively as recurring concepts and perspectives that represent experiences across multiple participants. Themes also emerged from informal comparisons of positive/negative perspectives across intervention types. Based on prior work, we expected themes to emerge about the “realness” of the exposures (i.e., the extent to which it felt like feared events in real life), “accessibility” of the exposure (i.e., ease of accessing the exposure), and “novelty” particularly related to the technological novelty of using VR in particular.

2. Results

2.1. Baseline differences by condition

No significant differences were found between homework condition groups in demographics (age or gender assigned at birth) or pre-clinical assessments (see Table 1). Due to small sample sizes in some categories, statistical tests comparing homework condition groups on race, ethnicity, and current identified gender could not be conducted.

2.2. Hypothesis #1: effect of in-session exposure type on anxiety ratings

As illustrated in Fig. 1, longitudinal models of Session 1 and Session 2 SUDS ratings using last observation carried forward (LOCF) demonstrated that average SUDS ratings significantly decreased across exposures for both exposure types during both study sessions. Model estimated average change in SUDS rating across 10 ratings were as follows: Session 1 VRE = −0.40, F(1, 449) = 140.9, SE = 0.03, and IE = −0.41, F(1, 449) = 168.0, SE = 0.03; Session 2 VRE = −0.23, F(1, 409) = 78.8, SE = .02, and IE = −0.23, F(1, 409) = 94.3, SE = 0.02, all four p < .0001. In addition, the SUDS initial and maximum ratings decreased from Session 1 to Session 2 for both exposure types (SUDS initial average change VRE=−3.2, t(40) = 8.3, IE= −3.0, t(40) = 7.4, all p < .0001; SUDS max average change VRE= −3.4, t(40) = 8.5, IE= −3.1, t(40) = 8.1, all p < .0001). Contrary to the hypothesis that VRE would result in higher SUDS ratings, the average difference in SUDS ratings between the two exposure types (0.4 higher for VRE) was not statistically significant (p = 0.37). The results were the same using all available data as with LOCF. Paired sample t-tests (Table 3) did not find significant differences across exposure type for SUDS ratings at various time points (initial, maximum, final) nor for decrease in SUDS over the course of the exposures. VRE lasted longer in minutes than did verbal IE in both sessions, but the two types of exposure did not differ in number of ratings to exposure completion (Table 4).

Fig. 1.

Average Anxiety (SUDs) Ratings by Exposure Type and Study Session. SUDS = Subjective Unit of Distress; i.e., participant’s rating of anxiety in the moment. LOCF = Last observation carried forward. Because of the discontinuation rules, participants were not required to continue with exposures through 10 ratings. As such, the last SUDS rating provided was carried forward. VR = Virtual Reality Exposures. IE = Imaginal exposure completed with paper and pencil.

Table 3.

Comparison of exposure type on anxiety ratings (SUDS) and duration measured in minutes and number of SUDS ratings, using paired t-tests.

	Exposure Type
	IE	VR	Difference
Session 1 (n = 45)	Mean (SD)	Mean (SD)	Mean (SD)	95 % CI	t-value	p-value
First Anxiety Rating	5.2 (2.3)	5.0 (2.1)	−0.2 (2.2)	−0.9, 0.5	−0.54	0.5903
Maximum Anxiety Rating	5.2 (2.1)	5.4 (2.4)	−0.2 (2.2)	−0.9, 0.5	−0.60	0.5492
Final Anxiety Rating	1.6 (1.3)	1.6 (1.4)	0.0 (1.4)	−0.5, 0.4	−0.22	0.8301
Decrease in Anxiety Rating*	3.7 (2.0)	3.6 (1.9)	−0.1 (2.0)	−0.8, 0.5	−0.52	0.6024
Exposure duration (min)	10.8 (5.7)	14.7 (9.2)	−3.9 (9.3)	−6.7, −1.1	−2.81	0.0074
Number of ratings	4.9 (2.5)	5.4 (2.6)	−0.4 (0.4)	−1.3, 0.4	−1.06	0.2929
Session 2 (n = 41)
First Anxiety Rating	2.2 (2.1)	2.0 (1.9)	0.2 (2.0)	−0.5, 0.8	0.54	0.5943
Maximum Anxiety Rating	2.2 (2.1)	2.0 (1.9)	0.2 (2.0)	−0.5, 0.8	0.54	0.5943
Final Anxiety Rating	0.8 (0.9)	0.7 (0.9)	0.1 (0.6)	−0.2, 0.2	0.50	0.6230
Decrease in Anxiety Rating*	1.4 (1.7)	1.3 (1.4)	0.1 (1.9)	−0.5, 0.7	0.41	0.6853
Exposure duration (min)	3.1 (1.9)	7.5 (4.3)	−4.4 (4.3)	−5.8, −3.1	−6.61	< .0001
Number of ratings	3.3 (2.0)	3.1 (1.7)	0.1 (1.7)	−0.4, 0.7	0.45	0.6577

^*Decrease is calculated as first – last anxiety rating

Table 4.

Maximum SUDS during Session 2 cross-over^* xxposure and in vivo exposure.

		HW IE/VR in Session 2 Mean (SD)	HW VR/ IE in Session 2 Mean (SD)		p-value
Max SUDS T2 (crossover*)-		2.5 (1.8)		2.1 (2.3)	0.5741
Max SUDS T2 (crossover*)- adj for time 1					0.4819
Max SUDS Generalized Exposure T2		5.5 (2.6)		4.9 (2.8)	0.4941
Collected up to 30 times	Median (range)	5.0 (2, 10)	(1, 10)	50.

Note. SUDS = subjective units of distress scale.

^*Crossover exposure refers to the Session 2 VR exposure for the IE homework condition and to Session 2 IE exposure for the VRE homework condition.

2.3. Hypothesis #2: homework engagement

No significant differences emerged between homework conditions in number of homework exposures completed (VRE M = 5.4, SD = 1.5 vs. IE M = 5.9, SD =1.3), t(38) = 1.01, p = .32, children’s ratings of exposure homework enjoyment (VRE M = 5.4, SD = 1.9 vs. IE M = 5.1, SD =2.4), t(39) = −.38, p = .70, and parents’ ratings of the ease of getting children to complete exposure homework (VRE M = 7.6, SD = 1.7 vs. IE M = 6.5, SD =2.2), t(39) = −1.85, p = .07.

2.4. Hypothesis #3: preparatory effect of exposures

First, there were no significant differences in SUDS at various time points (first, maximum, and final SUDS and first-to-last decrease in SUDS) or exposure duration when comparing groups who completed the VRE first compared to those who completed verbal IE first when tested using two-sample t-tests as well as longitudinal (repeated measures) mixed models adjusted for rating number (Supplementary Table A). Second, when examining the effect of homework condition on Session 2 exposures, there was no difference between groups on max SUDS ratings during the cross-over exposure, i.e., Session 2 IE for participants in the VRE homework condition (mean = 2.1, SD = 2.3) vs. Session 2 VRE for participants in the IE homework condition (mean = 2.5, SD = 1.8), t(38) = .57, p = .57. Likewise, there was no difference in max SUDS ratings during the generalization in vivo exposure between patients that completed homework with IE (mean = 5.5, SD = 2.6) vs. those who completed homework with VRE (mean = 4.9, SD = 2.8), t(df) = X, p = .49. Results were similar with the additional multivariable linear regression model with a covariate adjusting for Session 1 maximum SUDs (p = .48).

2.5. Simulation sickness symptoms

Physical side effects as measured by the SSQ-m were, on average over both sessions, one point greater following VRE compared to IE (model based effect of 1.1, F(1,43)= 9.8, SE= .36, p = 0.005). There was also an overall effect for exposure order such that symptoms significantly decreased with subsequent exposures relative to the first exposure (model-based effects ranging from −0.8 for 2nd exposure to −4.2 for 4th exposure, F(3,114)= 8.0, over-all p < .0001). The model showed no significant effect for homework condition on SSQ-m scores (model based estimate = −.02, F(1, 43)= 0.01, SE= 1.67, p = .99).

2.6. User experience

Youth responses to the question asking which exposure type was more realistic were varied and, on average, favored VR exposures as more realistic than verbal IE, M = 7.05, SD = 2.36, Median = 8, min = 1, max = 10, t(40) = 5.56, p < .0001. No significant differences emerged between homework conditions in children’s ratings of exposure homework enjoyment (VRE M = 5.5, SD = 1.9, verbal IE M = 5.1, SD = 2.4, p = .70) and parents’ ratings of ease getting children to complete exposure homework (VRE M = 7.6, SD = 1.7, verbal IE M = 6.5, SD = 2.2, p = .07.

Interviews with families led to four themes: unrealistic (both), diminishing novelty (VRE), accessibility (IE, VRE with caveats), and being easily tailored to the individual’s fears (IE). Sample quotes are presented in Table 5. Whereas a theme of realism was expected, the theme that emerged reflected participants comments that both types of exposures seemed unrealistic, or less realistic than direct experience. Regarding VR, participants and their parents described the exposures as unrealistic in two ways: 1) the scenes sometimes felt inaccurate (e.g., a cafeteria scene looks different from their school cafeteria); and 2) the participant was acutely aware that they were completing a VR exposure that was not actually happening. Similarly, some youth reported awareness that the verbal IE was “not real.” Whereas a theme about the novelty of VR was expected, participants often described VRE as “cool” but reported that the novelty wore off quickly. The theme of novelty did not arise in discussing IE. A theme of accessibility was expected, and comments about accessibility were mixed for VR and positive for IE. Specifically, participants noted that VR exposures were more accessible compared to arranging for specialized exposures like blood draws or shots, but less accessible and difficult to manage than other effective exposure material (e.g., online videos). Families also experienced technical difficulties with VRE (e.g., device not charging). On the other hand, verbal IE was often described as very portable and required little advanced planning. The theme of individual tailoring was not expected a priori. Regarding this theme, participants reported that IE could be very detailed and specific to their own situations and could be adjusted over time. The theme of tailoring did not arise regarding VRE.

Table 5.

Themes and illustrative quotations regarding participant experiences with virtual reality and imaginal exposures.

Theme	Quotation
Unrealistic	Youth about VR: “I don’t think it’s the same, like, at all, ’cause I was able to like tell that like it wasn’t real, and like it didn’t matter, ’cause like those people couldn’t formulate opinions on me.” Youth about IE: “It just didn’t really feel real. … I just knew I was reading a paper.”
Diminishing Novelty	Youth about VR: “I think VR is really cool, and it was a—it was an interesting change to what I usually do. Youth about VR: “Once I do it one time, and then, like, that’s kinda cool, but then once you do it over and over and over again it gets really boring, and you feel like it’s kind of a waste of time.”
Accessibility	Parent about VR: “I think just having access to it and that making sure it’s charged would be… maybe a barrier… It’s really easy to just sit down and have her read the story.” Youth about IE: “Thought exposure is really something that you can do, like, anytime.”
Tailoring	Parent about IE: “She’s able to really go right into the specific of what she wants, whereas I think the, um, VR seems—it was the same thing over and over again. She—it didn’t seem like she was as challenged.” Youth about IE:: “I think it’s a good idea, and then it’s also easy to, like, add something new, I guess, if you’re, like, anxious about a different thing after a while.

3. Discussion

The current study demonstrated that both VRE and verbal IE functioned as expected in eliciting anxiety that diminishes with repetition but did not demonstrate superiority of one form over the other in evoking anxiety, homework completion, or preparing participants for subsequent exposures including in vivo exposures. Despite the promise of new technology to make VRE immersive and ratings of VRE as more realistic than verbal IE, these results did not support expectations that participants would experience VRE as more anxiety-provoking or engaging and thus more effective in preparing them for subsequent exposures. Finally, although of unlikely clinical significance, VRE was associated with more physical symptom side-effects than verbal IE. Overall, although we cannot prove a null result, our findings did not demonstrate expected advantages of VRE over verbal IE, even with technological advances. We discuss specific points related to exposure immersion and engagement below with particular attention to placing current study findings in context of past research and implications for future research and clinical application.

The current study builds on prior research demonstrating VRE as a feasible form of exposure in the treatment of anxiety disorders among youth (Blanco et al., 2023) and adults (Carl et al., 2019) through demonstration of feasibility of both VRE and verbal IE for a number of difficult-to-create feared scenarios in the treatment of childhood anxiety disorders. The inclusion of verbal IE as a comparison condition addressed a limitation of previous research that compared VRE to an inactive condition on treatment outcomes. The inclusion of verbal IE is additionally important because it is a relatively easy, low-tech option that has received little empirical attention (Whiteside et al., 2020). It is important to note that, whereas prior research examined a course of VRE treatment, our study was designed for a focused comparison of VRE and verbal IE over a limited time frame while youth participated in our clinic’s usual course of exposure therapy. Thus, although the current findings can be interpreted as supporting the use of either and both forms of imaginal exposure in the treatment of childhood anxiety disorders, they should not be interpreted as VRE and verbal IE having similar effects on treatment outcomes. Although such a study could be designed, we argue it would not be worthwhile for a number of reasons. First, multiple studies have now found no differences in VRE and verbal IE, this one and one previous (Whiteside et al., 2020). Second, participant feedback that both forms of imaginal exposure were limited in their realism underscores expert opinion on the importance of including in vivo exposure whenever possible in a course of treatment (Whiteside et al., 2020). Third, our findings indicate that both forms prepared participants for a related in vivo exposure. Taken together, clinical implications of our findings are that both VRE and verbal IE could be effective and useful tools within a course of exposure therapy that leads to facing fears in real life.

The current study also expands on our previous research comparing VRE and verbal IE (Whiteside et al., 2020) by providing the strongest comparison to date. Contrary to expectations that improved, immersive VR technology would make for more realistic and, thus, anxiety-provoking VRE content, VRE did not demonstrate expected advantage over verbal IE in terms of eliciting anxiety or in preparing participants for subsequent exposures. These findings are consistent with our prior study (Whiteside et al., 2020) that found VRE and verbal IE did not differ in their elicitation of anxiety and extends those findings by demonstrating the two forms prepared participants for in vivo exposures and through inclusion of a broader array of content. Specifically, whereas the prior study included only exposure to academic performance worries, the current study included multiple scenarios that are difficult to create in office such as school settings, interactions with peers and teachers, presentations to peers, and needle procedures in a medical office.

Although the current study is not designed to established equivalence, it is unlikely that the shortfall of VRE can be attributed simply to methodological shortcomings. To begin with, the experimental manipulation of eliciting anxiety that dissipates with repetition was successful. The generally moderate levels of initial anxiety that declined to low levels of anxiety over 10–15 minutes is consistent with data on exposures during clinical treatment (Whiteside et al., 2019). Moreover, the patients were youth with clinically diagnosed anxiety disorders of a severity comparable to the clinical population from which they were selected (Brennan et al., 2022). As such, it appears that the experimental design successfully replicated clinical conditions. The most likely contributor to the inability of VRE to outperform IE in terms of eliciting anxiety is the failure of either format to achieve a heightened immersive realistic experience. Although qualitative ratings of realism favored VRE slightly, participant comments noted being aware that the VR exposure was not in fact reality and that their actions in response to the VR scenario would not lead to any consequences. These limits to full engagement may have prevented the VR exposure from sufficiently establishing the parameters to disprove a child’s anxiety-provoking expectations. The failure of VRE to demonstrate advantage in terms of participant engagement might have been influenced by the intentional design choice for participants to repeat one VRE and one IE during the study. Whereas this decision was intentional for experimental control, a design in which participants have a menu of exposures to complete for VRE and IE could lead to different results in terms of self-directed exposure completion and perceptions of novelty. Such design changes would be unlikely to alter participants’ experience of VRE as sometimes cumbersome (e.g., issues with charging the batteries) and their perceptions of verbal IE as easily portable and capable of individual tailoring.

Although the current null results are consistent with our earlier study, they run counter to findings in the adult literature in which VRE has been of similar effectiveness to in vivo exposure (Carl et al., 2019). One potential explanation may be the combination of multiple scenes and fears (e.g., social anxiety and needle phobia) in our study, which could potentially obscure the relative advantage of VRE over IE in a specific area. Qualitative themes and participant comments suggest relative applicability of VRE and verbal IE depending on the individual and situation. For example, VRE may provide fitting exposures when the details in the content align with the youth’s experience and/or feared outcomes. Future research might explore this possibility, although the lack of evidence supporting the idea that VRE is more powerful than IE and previous adult literature suggesting that the number of sessions of VRE use does not relate to the effectiveness of VRE (Carl et al., 2019), provides little support for this endeavor. Perhaps such a study would be beneficial for other, more difficult to replicate fears, such as fear of flying, fear of driving, and traumatic events in the context of PTSD; VR exposures for such fears have been shown to be more effective compared to waitlist or placebo conditions among adults and, in the case of specific phobias, equally effective to in vivo exposure (Carl et al., 2019); however, fears of flying and driving may not be common enough in children, rendering application for this population low yield, and variability in traumatic experiences of youth (in contrast to combat experience in adult military personnel) would complicate development of applicable content to treat PTSD in children. Furthermore, studies examining whether VRE helps increase the effectiveness of self-guided exposures may be of clinical relevance as they could allow for fewer clinician hours in patient care (e.g., Zainal et al., 2021).

3.1. Limitations and future directions

The current study was not powered to test equivalence; however, the observed mean differences in this study were small and any statistically significant difference in a more highly powered study is unlikely to be clinically meaningful. Participants in the current study received just one VRE and one verbal IE to practice repeatedly for homework between study sessions, which might have influenced participants’ perceptions of diminishing novelty. The distribution of participants across exposure content areas did not allow examination of effects by content (e.g., social vs needle) or degree of interaction in the exposure (e.g., watching only vs. voice-activated interaction). Further study with explicit comparisons would be needed to explore the potential of specific types of VRE and/or verbal IE. In addition, creation of more content and opportunities to “level up” (see Zainal et al., 2021) might better sustain novelty and engagement. Finally, the current study did not quantify costs and relative value of VRE vs. IE; however, given the lack of significant differences in this study and significant cost differences of paper and pencil compared with a VR headset and creation of VRE content, a clear advantage of VRE over verbal IE would be needed to justify its financial value or cost-effectiveness.

4. Conclusions

The current study expands upon the previous literature on VRE for CADs be comparing a more varied and technological advance version to verbal IE. Consistent with earlier research with older, presumably less realistic VRE content, results of this study did not demonstrate any clear advantages of one form over the other. Considering likely differences in costs of VRE equipment and content development compared to paper and pencil and the current pediatric research evidence, it is difficult to make a case of added value of VRE over verbal IE in youth with pediatric anxiety.

Appendix A. Supporting information

Supplementary data associated with this article can be found in the online version at doi:10.1016/j.janxdis.2025.102974.