Cognitive behavioural therapy in virtual reality treatments across mental health conditions: a systematic review

Background. Virtual reality (VR) has been effectively used in the treatment of many mental health disorders.However, significant gaps exist in the literature. There is no treatment framework for researchers to use when developing new VR treatments. One recommended treatment across a range of diagnoses, which may be suitable for use in VR treatments, is Cognitive Behavioural Therapy (CBT). The aim of this systematic review is to investigateCBT treatment methods that utilize VR to treat mental health disorders. Objectives. To investigate how CBT has been used in VR to treat mental health disorders and to report onthe treatment characteristics (number of sessions, duration, and frequency) that are linked to effective and ineffective trials. Methods. Studies were included if patients had a mental health diagnosis and their treatment included immersiveVR technology and CBT principles. Data were extracted in relation to treatment characteristics and outcomes,and analysed using narrative synthesis. Results. Ninety-three studies were analysed. Exposure-based VR treatments were mainly used to treat anxiety related disorders. Treatments generally consisted of eight sessions, once a week for approximately one hour. VR treatments were commonly equal to or more effective than traditional face-to-face methods. No specific treatment characteristics were linked to this effectiveness. Conclusion. The number, frequency and duration of the VR treatment sessions identified in this review, could be used as a treatment framework by researchers and clinicians. This could potentially save researchers time and money when developing new interventions.


INTRODUCTION
Virtual reality (VR) is a technological interface that allows users to experience computer-generated environments within a controlled setting [1]. Recent meta-analyses and systematic reviews have found this technology to be an effective tool in the treatment of a range of mental health conditions [2], with most evidence derived from anxiety-related disorders [3], eating disorders [4] and psychosis [5].
In addition to its treatment effectiveness, VR exposure therapy has been found to be more cost-effective than face-to-face treatment for posttraumatic stress disorder [6]. Furthermore, VR treatments are well accepted by patients, who have expressed high levels of support and interest in its use for their mental health treatment [7].
There is also evidence that drop-out rates may be lower with VR treatments than with traditional face-to-face treatments [8]. This technology may, therefore, potentially improve access and adherence to psychological treatments [7,8].
Despite the potential of VR in mental health treatment, significant gaps exist in the literature relating to VR treatment. Studies in the literature have mainly focused on treating anxiety disorders with exposure-based therapies and have overlooked other diagnoses (e.g., depression, bipolar and personality disorder) and other treatment possibilities (e.g., guided self-help) [8].
A framework is a basic structure that underlies a system or concept, and may be built on or used as a point of reference to decide upon a particular course of action [9]. To our knowledge, there are no shared VR treatment frameworks currently available for researchers to follow. Without a treatment framework on which to build, researchers who want to explore new VR treatment methods for overlooked diagnoses, are forced to spend a great deal of time and money to develop their own treatments, which may or may not при разработке новых методов ВР-терапии. Одним из рекомендованных методов терапии при широком спектре диагнозов, который может применяться в формате ВР, является когнитивно-поведенческая терапия.
One recommended treatment across a range of diagnoses [11], which may be suitable for use in VR treatments, is cognitive behavioural therapy (CBT). CBT is based on the cognitive model of mental illness and this model hypothesizes that the way in which patients feel and behave, is determined by their perception of situations, rather than the actual situations [12].
CBT aims to relieve distress by helping patients develop more adaptive cognitions and behaviours [13].
Developing a treatment framework that summarizes effective VR CBT treatment characteristics (e.g., the number of sessions, duration and frequency) could provide a possible foundation upon which researchers can build. This could potentially reduce the time and money spent on the development of interventions.
At present, no research has synthesized VR treatment characteristics across diagnoses. The aim of this systematic review is to explore CBT treatment methods that utilize VR to treat mental health disorders.
A treatment framework will be developed from the identified shared treatment characteristics (e.g., the number of sessions, duration and frequency).

Objectives
The objectives of this systematic review are to: 1 investigate how CBT has been used in VR to treat mental health disorders.
2 report on the treatment characteristics (number of sessions, duration, and frequency) that are linked to effective and ineffective trials.

METHODS
The study protocol for this systematic review and narrative synthesis was registered on PROSPERO

Identification of studies
The eligibility criteria were developed using the PICO framework [14]. Papers were eligible if they were written in English, the study participants had to be over the age of 18 with any mental health diagnosis, using recognized diagnostic criteria (ICD-10 or DSM-V) or a validated scale with a pre-defined cut off point.
To be included in the review, the interventions in the studies had to use principles of CBT, as defined by the NHS [15]. Furthermore, the VR technology used, had to be immersive. Immersive VR is defined as a computersynthesized virtual environment surrounding the user.
This can include (but is not restricted to) a headmounted display (HMD) and a Cave automatic virtual environment (CAVE). An HMD consists of a computergenerated video display attached to the user's head, with retina or head trackers that measure the changing position, which is fed back to the rendering computer [16]. A CAVE is essentially a room in which computergenerated visual imagery is projected onto the walls, floor and ceiling, and the user is free to move around [17]. Papers were excluded if they did not have an experimental design (e.g., case series and reviews) and if the treatment procedures were not reported.  [8] and was further developed in conjunction with an information scientist. The general search terms were: 'virtual reality' AND 'cognitive behavioural therapy' AND disorder-specific terms (see Appendix A for full search terms). Databases were searched from inception for titles, abstracts and keywords. Four key papers were identified and used to assess the reliability of the search results [1,8,18,19]. The authors also conducted hand searches of the Annual Review of CyberTherapy and Telemedicine and the reference list of relevant papers. Study authors were contacted when access issues occurred.

Study selection
Identified references were transferred into Endnote and duplicates removed. The references were then transferred into an Excel spreadsheet. The first reviewer (MD) screened all the titles and abstracts, whereas the second reviewer (NL) independently screened 25%. Subsequently, the full text of the potentially relevant papers was retrieved and was once again independently assessed for eligibility by both reviewers. Hereafter, the reasons for exclusion were noted in the database. The inter-rater reliability for screening between the authors (MD and NL) using Cohen's Kappa was moderate (60% agreement, p<.0001). Any disagreements throughout the screening process were resolved through discussion and, if necessary, by involving a third reviewer (VB).
A data extraction framework was created using Excel and piloted with five studies. The data extracted included general information relating to the study eligibility, methods, VR treatment descriptions and a summary of the results, outcomes and conclusions.
Some treatment characteristics such as number and duration of sessions, were reported numerically, other treatment characteristics such as type of VR technology used and treatment location, were simplified into categorical variables for quantitative synthesis. This was to allow synthesis and integration of a large amount of data across the dataset. The quantitative data were imported into SPSS to allow for vote counting and for the statistical testing of differences. Vote counting and quantitative synthesis (e.g., t-tests and Chi-squared) were used to develop a preliminary synthesis, as they allowed the researchers to identify patterns across the included studies [20].
The first objective of this review was to investigate how CBT has been used in VR to treat mental health disorders. Once the treatment characteristics of all of the 93 studies were synthesized, the first objective of this review had been achieved.
The second objective was to report on the treatment characteristics that are linked to effective and ineffective trials. Studies which aimed to explore VR treatment effectiveness (62 out of the 93 studies) were selected for the second analysis. These studies were categorized according to their aims and were analysed separately.
Finally, treatment characteristics of studies which found VR to be more effective by comparison with 'traditional' treatment methods (e.g., in-vivo exposure) were compared with studies which found VR to be ineffective when compared with 'traditional' treatment methods. To allow for a clear comparison, studies which were equally effective using 'traditional' methods were excluded from this analysis.

Risk of bias
The two reviewers (MD and NL) independently assessed the risk of bias using the Quality Assessment Tool for Quantitative Studies [21]. This tool has been specially developed for public health research and assesses six components of bias and quality; these include selection bias, study design, confounders, blinding, data collection methods, withdrawals and dropouts. The inter-rater reliability between the authors, using Cohen's Kappa was high (80% agreement, p<.001). Any disagreements between the two reviewers were resolved through discussion or by consulting a third reviewer (JM).
The results of the quality analysis were further tabulated to identify any types of bias common to the included studies.

RESULTS
The study selection process and a summary of the included studies will be presented first, followed by a general overview of the quality of the included studies.
Next the main results will be presented, according to the two review objectives; 1) how CBT has been used in VR to treat mental health disorders and 2) which are the treatment characteristics that are linked to effective and ineffective trials.

Selection and inclusion of studies
Once duplicates were removed, the search generated

Excluded (n =162)
VR is not immersive n=45 No treatment procedures reported n=28 Not an empirical design n= 28 No CBT principles n= 23 Study not on mental health treatment n=16 Mental health diagnosis is not predetermined n=11 Trial report unavailable after contacting authors n= 4 No VR treatment n=4 Participants have cognitive disorder n=1 Participants younger than 18 years old n=1 Study not in English n=1

Quality assessment and risk of bias in included studies
The quality of the studies in this review was found to be predominately weak (see Appendix B for individual study quality assessment). A table was formulated to explore why studies were often of poor quality ( Table 1).
The poor quality of the studies can be attributed to selection bias. Most studies either did not report where they recruited their patients from, or they recruited volunteers through advertising. This may have resulted in lower than anticipated drop-out rates, as volunteers might have been more willing to participate. Furthermore, although logistically difficult, most studies did not blind the patients or the assessors to the treatment intervention.
This may have resulted in assessment bias.
A cross-tabulation between the quality of studies and the year of publication showed that the quality of studies has not improved over time.

How has CBT been used in VR to treat mental health disorders?
To address the primary aim of the review, the common characteristics of treatments will be described. For a summary of the treatment characteristics, please view the second column of Table 2.
VR has generally been used as a component in a more extensive treatment protocol (n=58). On average, patients were offered eight treatment sessions, and six of these sessions involved VR technology. The first and the last sessions were psychoeducational, e.g., identifying symptoms and discussing relapse prevention [22]. During VRET, patients are gradually exposed to a virtual environment that provokes anxiety, e.g., a battlefield in the case of patients with post-traumatic stress disorder [23], or exposure to a spider for patients with arachnophobia [22]. The aim is that patients become desensitized to the fear-provoking stimuli with gradual exposure.
VRET was the most commonly used treatment in this review. Across the nine remaining studies, there was some variation in the definitions used to describe the CBT treatments, e.g., VR enhanced CBT, VR cognitive therapy and repeated behavioural experiment tests.
These treatments will be discussed together. Similar to VRET, these treatments all used VR to expose patients to specific, anxiety-provoking virtual environments.
However, unlike VRET, the aim of exposure was not just to desensitize the patient to a situation, but to trigger certain emotions or behaviours that therapists can subsequently work on with the patient. For instance, in an eating disorder study, patients were exposed to virtual environments that were thought to trigger emotions related to weight, e.g., restaurants, clothes shopping and a swimming pool. In these environments, patients performed virtual tasks such as weighing themselves and trying on clothes, whilst the therapist discussed feelings and beliefs [24]. Similarly, Pot-Kolder et al. (2018) [26] used VR to expose patients with persecutory delusions and paranoid ideation to stressful social environments, that could trigger fear and paranoid thoughts, e.g., being on the underground or in a café. In these virtual environments, they explored and challenged the patient's suspicious thoughts and safety behaviours, and tested harm expectancies.
One study [27] 53  provide an overview of all the aims (n=93) then it will specifically focus on the subgroup of studies that aimed to investigate treatment effectiveness (n=62).

Duration of VR sessions (in mins)
The aim of the papers correlated with the year of publication, demonstrating that earlier studies tended to focus on assessing the efficacy of VR treatments, whereas later studies aimed to assess the use of cheaper technology and remote treatment delivery. For a summary of the study aims, please see Table 3.

Efficacy of VR treatment within a specific patient population
Of the 20 studies that used a repeated measures design to investigate the efficacy of VR treatment with a specific patient population, all considered VR to be an effective treatment for anxiety-related disorders [119][120][121], substance disorders [31] and eating disorders [122].
For instance, a cohort study comprising 20 combatrelated PTSD patients reported post-intervention, that following VRET, 80% of the patients no longer met the criteria for PTSD [23]. A breakdown of the treatment characteristics in studies that found VR treatment effective within a specific population, can be found in Table 2. These studies generally consisted of a small sample size (M=25.4, SD=26.8, n=20). The treatments involved a mean of nine sessions, and VR was used in seven of these sessions. The treatment was delivered once a week for a mean duration of 54 minutes.

Efficacy of VR treatment by comparison with waiting list
Similar to studies that investigated the effectiveness of VR treatment within a specific patient population, the majority of the studies reported VR treatments to be relatively more effective than waiting list controls (n=7).
A controlled clinical trial with 23 arachnophobia patients, reported that VRET was effective in treating this phobia.
Eighty-three per cent of the patients in the VRET group showed a significant clinical improvement by comparison with no improvement in the waiting list group [49]. An RCT, with 116 psychotic disorder patients, found that VR-CBT did not increase the length of time patients spent with other people, however, it did significantly improve patients' momentary paranoid ideation and anxiety.
These improvements were maintained six months after completion of follow-up treatments [26].
Only one RCT that had 32 general anxiety disorder patients, reported that a single session of VRET was not significantly effective by comparison with the waiting list group [54].
The fourth column of Table 2   were delivered across a mean of 10 sessions, and seven of these involved VR.

Effectiveness of VR treatment by comparison with 'traditional' treatment methods
The majority of the studies in this review aimed to identify the effectiveness of VR treatments by comparison with 'traditional' treatment methods.
Thirty-one out of the 34 studies (91.2%) considered VR treatments to be equally or more efficacious than traditional treatment methods. See Table 4  diaries [130]. The patients' reasons for dropping out of 'traditional' treatments included not wanting in-vivo exposure [71], not being satisfied with the treatment allocation and wanting to pay for VR therapy [69]. such as the number and duration of the sessions, were very similar across the two outcomes. Please see Table   2, Column 6 for comparisons between the variables.

Main findings
VR has mainly been used in the treatment of anxiety- Even though the overall quality of the evidence is weak, VR treatments seemed to perform comparably in terms of efficacy with 'traditional' face-to-face treatments.
Treatment characteristics, such as the number and duration of sessions, were very similar between studies that regarded VR treatment as effective and those that found it not to be effective. However, patient drop-out rates were significantly lower in studies that considered VR treatment to be effective by comparison with those that found it ineffective.

Comparison with literature
This review is the first to investigate how VR has been used in CBT (a psychotherapeutic approach) to treat a variety of mental health disorders. Previous VR reviews have focused on providing a general overview of the field [8] or reported treatment outcomes for specific diagnoses [2].
Results from this review support the findings from previous reviews, that VR is an acceptable and promising therapeutic tool for mental health treatment [4]. It can be used to deliver cognitive rehabilitation, social skills training interventions and VR-assisted therapies for psychosis [5]. VRET is equally effective as in-vivo exposure for the treatment of anxietyrelated disorders [3]. is also unable to make any firm conclusions regarding these differences, but it does highlight the importance of offering high-quality treatments in research studies.

Strengths and Limitations
This review is the first to collate data as to how CBT has been used in VR to treat mental health disorders. to all the data in the study and took the final decision to submit for publication.

Conflict of Interest:
The authors report no conflict of interest.