In this paper, we use the term autistic adults, in line with international recommendations [1] to refer to adults on the autism spectrum. Additionally, we use the term autism spectrum condition as an equivalent to autism spectrum disorder to minimize discriminatory language.

According to the diagnostic criteria described in the DSM-5-TR (Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision) by the American Psychiatric Association, autism is defined as a neurodevelopmental condition characterized by persistent deficits in social communication, such as difficulties with socioemotional reciprocity, communication, and the development, maintenance, and understanding of relationships. This is accompanied by restricted, repetitive patterns of behavior, interest, or activities. These clinical characteristics must be present from early development (but may not become fully manifest until social demands exceed limited capacities or may be masked by learned strategies in later life), cause significant impairment in psychosocial functioning, and cannot be better explained by other intellectual or developmental conditions [2]. Autism is a lifelong neurodevelopmental variation that is estimated to occur in 1%‐2% of the global population [3]. The prevalence rate is increasing [4,5], as well as the lifelong economic impact associated with the condition. In 2019, the estimated lifetime social cost of an individual on the autism spectrum was approximately US $3.6 million in the United States, with the largest costs associated with lost productivity and adult care [6]. These costs should not be attributed to individuals on the autism spectrum but should rather be regarded as a consequence of inadequate support and accommodations in educational and occupational settings and a lack of effective interventions that address the specific challenges faced by the autistic population. These obstacles are particularly pronounced for autistic adults, as many experience a lack of appropriate interventions and support [7], are vastly underrepresented in the workforce, and are at heightened risk of depression and suicidal actions compared to the general population [8-11]. Despite the critical need for interventions and support systems aimed at improving outcomes for autistic adults, research on autism has primarily focused on children and adolescents, overlooking the growing population of adults who face extensive unmet treatment needs [12].

Social cognition can be defined as “(…) the mental operations that underlie social interactions, including perceiving, interpreting, and generating responses to the intentions, dispositions, and behaviors of others.” [13]. As this definition implies, social cognition is a multifaceted and complex construct, underpinned by different functions, such as emotion processing and recognition, theory of mind (ToM), empathy, social perception, social attribution, and social awareness. It is well established that autistic adults, on average, show reduced performance on social cognitive tasks, including facial emotion processing, prosody interpretation, mental state labeling, social perception, social attribution, and social understanding, compared to neurotypicals. However, social cognitive profiles in autism present differently from person to person, and individual profiles can be dynamic and may shift over time [14].

Social cognitive impairments are key deficits faced by autistic adults [15], which severely impact social interactions, hinder educational and vocational functioning, and compromise quality of life [16-18]. The social cognitive deficits in autism encompass both the more basic functions, such as recognizing emotions or making appropriate eye contact, and higher-order functions, such as inferring others’ mental states (ToM) or understanding subtle aspects of social communication [19,20]. These functions are vital to understanding and reacting adequately to dynamic, complex social situations in all aspects of life, from making friends and building intimate relationships in private life to being able to participate appropriately in educational or professional settings.

Targeting social cognitive deficits in autistic adults without using virtual reality (VR) technology has been shown to improve psychosocial functioning [21]. These interventions often use computer- or group-based exercises to improve different aspects of social cognition (eg, perception of social cues, emotion recognition, ToM, and social skills) through repeated practice and skills-based training. While preliminary evidence supports the effectiveness of targeted non-VR-based social cognitive interventions for autistic adults [22], the existing studies are limited in number and are subject to significant methodological limitations, such as not being randomized clinical trials or having small sample sizes (n≤17) categorizing them as pilot studies. Previous studies of social cognition training for autistic adults typically target multiple domains, such as emotion processing and recognition, ToM, and social perception and attribution. Hypothesized mechanisms of change include repeated practice, gradual and errorless learning, compensatory strategy training, and supportive contexts that promote generalization [22-24]. Specifically for VR-based interventions, additional mechanisms of change include immersion in ecologically valid scenarios, graded exposure, individually tailored training, real-time feedback, control of sensory load, and structured transfer to daily life [25,26].

The use of VR holds great potential for mental health interventions by enabling researchers and clinicians to evoke psychological reactions to lifelike scenarios in real-time within a controlled laboratory environment. VR serves as a powerful tool for delivering exposure therapy, as it creates a safe, flexible, and controlled space for immersive exposure to challenging situations. This approach allows individuals to practice confronting real-world challenges without facing the consequences of real-life situations, while also reducing the social pressure often associated with face-to-face interactions [27]. As a result, VR exposure is considered an effective tool for treating various psychological challenges, with minimal risks of adverse side effects [28]. Additionally, creating in vivo exposure to challenging social situations in daily life is often difficult to arrange and nearly impossible to control and can be overly stressful for participants, making engagement in such settings challenging. Interventions using VR exposure can overcome these obstacles by offering complete control over various scenarios and their elements, leading to greater user acceptance. As a result, participants may be more inclined to engage in exposure therapy when it is provided in a VR setting [29]. Additionally, VR enables the creation of highly personalized interventions by allowing control over the level of exposure, tailoring scenarios to individual needs, and making real-time adjustments.

By using immersive VR exposure in interventions targeting social cognitive challenges in autistic adults, it becomes possible to assess real reactions and responses to life-like challenging social situations, allowing access to the real-time cognition of the participant in such situations. This approach might be of particular importance for interventions aimed at individuals on the autism spectrum, since self-awareness and self-insight are often compromised by alexithymia (difficulties in identifying and describing one’s own emotional state) [30]. Alexithymia can hinder the accurate assessment of psychosocial difficulties in interventions, as more traditional approaches often require individuals to describe their own emotions retrospectively. Therefore, using VR exposure might enhance assessment accuracy and, in turn, enable more personalized and effective interventions. Furthermore, individuals on the autism spectrum may prefer technology-based interventions [23], and preliminary evidence indicates that VR-based interventions lead to faster symptom reduction than traditional psychological interventions [31,32].

Given that social cognitive differences in autistic adults often involve emotion processing and recognition, ToM, social understanding, and motivation [14,15], VR-based interventions are theoretically well-suited to target these domains. The ability to simulate complex, dynamic social interactions in a controlled, repeatable, and safe environment enables targeted practice of interpreting and responding to social cues, adjusting difficulty as skills improve [24,33]. By providing immediate feedback, opportunities for graded exposure, and the capacity to individualize scenarios, VR can support the gradual enhancement of underlying neuropsychological processes essential for real-world social functioning.

Research on VR-based interventions for individuals on the autism spectrum has gained significant momentum in recent years. Most previous studies have investigated interventions targeting specific functional domains, such as job interviewing skills, driving skills, and adaptive social skills [26,34]. While some studies have examined the impact of VR-based interventions on cognition in individuals on the autism spectrum, these studies have primarily addressed specific neurocognitive domains and daily life skills [35], with comparatively fewer studies investigating social cognition [24,25]. The results of these studies have provided preliminary evidence supporting the feasibility of VR-based interventions for individuals on the autism spectrum. These studies have yielded promising results, indicating positive effects on specific functional and social skills, as well as neurocognitive domains like attention, memory, and executive functioning and social cognitive domains including emotion recognition and ToM. However, the findings are limited by small sample sizes, lack of randomized, clinical trials, and insufficient long-term follow-up. Additionally, most research has focused on children and adolescents. Therefore, large-scale, methodologically rigorous trials are urgently needed to investigate the effectiveness of VR-based interventions for autistic adults to draw more firm conclusions.

The aim is to conduct the hitherto largest clinical randomized trial globally, comparing VR-based social cognitive training (VRSCT) + treatment as usual (TAU) to TAU for autistic adults. The objective is to examine whether VRSCT is effective in improving psychosocial functioning, quality of life, and co-occurring clinical symptoms and to reduce social cognitive deficits in autistic adults. Additionally, we will examine whether the VRSCT is cost-effective.

We hypothesize the following:

STEPS (Social Cognitive Training Enhancing Pro-Functional Skills) is a randomized, assessor-blinded, parallel-group superiority clinical trial. A total of 140 participants will be allocated to receive either VRSCT+TAU or TAU alone. Assessments will be carried out at baseline, at cessation of intervention (3 months post baseline), and at 6 months post baseline. A flow diagram of the trial is presented in the Results section. A stratified block randomization with a concealed randomization sequence will be conducted with a 1:1 allocation. Independent assessors, blinded to the treatment, will evaluate outcomes. The analysis of outcomes will be carried out according to the intention-to-treat principle [36].

The trial will enroll participants from public and private outpatient facilities in Denmark that specialize in the clinical assessment and diagnosis of autism in adulthood. The clinical staff at these sites, including nurses, psychologists, doctors, and social workers, will approach relevant participants and refer them to the project group. Additionally, relevant participants from community-based social support services in the Capital Region of Denmark will be referred to the trial by the multidisciplinary staff at these locations. Participants will also be recruited via TrialTree, an online recruitment platform that specializes in connecting interested participants with targeted and relevant clinical research projects. Furthermore, self-referrals will be accepted, provided that the eligibility criteria for participation are met. Assessment, randomization, experimental intervention, and analysis of outcomes will be carried out by the VIRTU Research Group located at Mental Health Center Copenhagen, 2900 Hellerup, Denmark.

Participants must provide written, informed consent prior to the initiation of any assessment or intervention in the STEPS trial. Inclusion and exclusion criteria are listed in Textbox 1.

The study is conducted in Denmark, where the ICD-10 (International Statistical Classification of Diseases and Related Health Problems, Tenth Revision [37], remains the current standard for diagnostic classification in clinical practice. As no autism diagnostic assessment will be conducted as part of the trial, participants are required to provide documentation confirming a clinical diagnosis of autism based on ICD-10 criteria (ICD-10 codes, namely F84.0, F84.1, F84.5, and F84.8). Diagnostic assessments based on autism-specific instruments, for example, Adult Asperger Assessment [38], Autism Diagnostic Observation Schedule, Second Edition (ADOS-2 [39]), and Autism Diagnostic Interview - Revised (ADI-R [40]), preferably supplemented by comprehensive developmental history and information from multiple informants (eg, self-report and informant-report) and from different modalities (eg, observation, interviews, and questionnaires), as recommended for improving autism diagnostic procedure, especially relevant for girls and women [41], are accepted. In cases where a diagnosis was confirmed without using formal assessment tools, documentation will be accepted only if the evaluation was conducted by an experienced clinician and is adequately documented.

To ensure participants experience significant deficiencies in social interactions, the Social Responsiveness Scale, Second Edition, Adult Form, Self-report (SRS-2A) will be used to assess the level of social impairments. A sex-normed T-score of ≥60 on the SRS-2A total score was chosen as the cutoff since a T-score of 60 is the lower threshold for mild deficiencies in social behaviors that are clinically significant and may lead to mild or moderate difficulties in daily social interactions [42]. Estimated full-scale IQ will be estimated using the 2 subtests, Vocabulary and Matrix Reasoning, from the Danish version of the Wechsler Adult Intelligence Scale, Fourth Edition (WAIS-IV) [43], which are known to correlate strongly with full-scale IQ [44]. Participants with co-occurring attention-deficit/hyperactivity disorder (ADHD) must demonstrate a preliminary satisfactory treatment response to ADHD medication at the time of enrollment. This determination will be based on mutual agreement between the participant and the clinician responsible for their ADHD treatment, confirming that targeted ADHD symptoms are well managed. Additionally, no changes in the type or dosage of ADHD medication are permitted within the 4 weeks preceding enrollment in the trial. The type and dosage of ADHD medicine will be thoroughly mapped and documented at enrollment, and at follow-up assessments. Further details of this criterion are displayed in the “Relevant Concomitant Care and Interventions Permitted or Prohibited During the Trial” section. Additionally, general psychopathology will be assessed with the Danish version of the Mini-International Neuropsychiatric Interview (MINI) [45].

Table 1 provides an overview of the outcome measures and their assessment time points.

The sample size calculation is based on detecting a between-group difference in the primary outcome, the SRS-2A, at cessation of the intervention (3 months post baseline). Previous studies have reported considerable variation in the SD of the social responsiveness scale total score when assessing social impairment in autistic adults [25,56,58]. As a result, we have not been able to determine the minimal clinically important difference. Therefore, we have based our sample size calculation on a medium effect size (Cohen d=0.5), which corresponds to a difference of 0.5 SD between the experimental group (VRSCT+TAU) and the comparison group (TAU). With 80% power and a 2-sided significance level of α=.05, 63 participants are required in each group to detect this difference. To account for an anticipated 10% dropout rate, we plan to recruit a total of 140 participants. Sample size calculation is based on a medium effect size (Cohen d=0.5) in accordance with established practice in clinical trial design. However, it may not reflect the smallest change considered clinically meaningful. Clinical significance will therefore be interpreted alongside observed effect sizes and CIs in subsequent reporting to ensure meaningful interpretation of trial outcomes.

The project group has established strong collaborations with several public and private outpatient facilities in Denmark, which are specializing in the clinical assessment and diagnosis of autism in adulthood. Based on the collaboration agreement, the multidisciplinary staff at these sites will routinely approach relevant participants and refer them to the VIRTU Research Group. To foster collaboration and ensure ongoing referrals to the trial, staff from the VIRTU Research Group will regularly attend conferences and meetings at the various recruitment sites and invite clinicians to discuss potential participants.

Data from evaluation of eligibility and assessments during the trial will be entered directly into the electronic case report form using the data entry system REDCAP. REDCAP is an electronic data capture tool hosted by the Center for IT, Medical Technology and Telephony (CIMT) in the Capital Region of Denmark. When electronic entry is not possible, data will be collected on paper and later transferred into REDCAP. Data on paper will be stored locally and secured. Using REDCAP, surveys can securely be distributed and returned via E-boks, a public Danish digital mailing system that all citizens possess, reducing the risk of data loss and unauthorized access. Data for each participant will be linked to a unique serial number, with only assigned researchers having access to REDCAP. REDCAP includes a complete audit trail of data transactions, detailed user rights, and access control management, ensuring compliance with Danish legislation (the Act on Processing Personal Data). Research data will be exported from REDCap without personal identifiers. Data will be exported to well-known software packages (eg, SPSS [IBM Corp], SAS [SAS Institute], Stata [StataCorp LLC], and R [R Foundation for Statistical Computing]) and placed in logged folders on a network drive under the control of the CIMT. A data manager will ensure that all variables are properly defined with variable and value labels. All derived variables will be properly defined, and algorithms will be kept in special files. All data will be examined carefully to identify errors in data entry.

Analyses will be performed on the intention-to-treat population for the primary and secondary end points, that is, analyzing all randomized participants according to the arm they were randomized to. The analysis method will be ANCOVA (analysis of covariance) of change from baseline with treatment as a factor and baseline value as a covariate at the corresponding postbaseline visit, that is, third month. Additionally, stratification variables and baseline values of social anxiety, alexithymia, and social function will be included as covariates. Treatment difference point estimates will be reported together with 95% CIs and P value. Missing data will be handled using multiple imputation. For the exploratory end points, a modified intention-to-treat set will be used, requiring at least 1 postbaseline measurement for the inclusion in the analysis for each end point. The analysis method will be the linear mixed effects model. Results will be reported for both postbaseline visits. Details of model specifications, imputation methods and assumptions, sensitivity analyses, and subgroup analyses will be provided in the forthcoming statistical analysis plan, if applicable. Results will be reported with a 95% CI and P value, using a 5% significance level. Baseline demographic and clinical characteristics will be summarized for each group, using descriptive statistics (eg, means, SDs, medians, and ranges for continuous variables and frequencies and proportions for categorical) to assess comparability between groups at baseline. Efficacy analysis will be conducted by a blinded and independent statistician with no contact with the participants in the trial. Reporting will follow CONSORT (Consolidated Standards of Reporting Trials) guidelines [85].

To address potential confounding from changes in psychotropic medication, we will conduct sensitivity analyses to explore whether medication changes influence the primary and secondary outcomes. Medication type and dosage are documented at both follow-up time points (3 and 6 months post baseline). To account for variability in TAU, we document the type, frequency, and duration of TAU received during two predefined periods (1) from enrollment to the 3-month follow-up and (2) from 3- to 6-month follow-up. If systematic between-group differences in TAU exposure emerge, summary TAU variables (eg, number of sessions and service type) will be included as covariates in a sensitivity analysis to adjust for potential confounding. Additionally, exploratory analyses may be conducted to examine whether specific aspects of TAU moderate the intervention effect. In addition, exploratory subgroup analyses will be conducted to examine whether selected baseline characteristics moderate the effect of the intervention on primary and secondary outcomes. Analyses will be based on several prespecified variables, including sex, alexithymia, social anxiety, depression, and ADHD.

Autistic adults constitute a growing and largely overlooked population, for whom few clinical and research resources have been allocated. Due to the limited evidence on effective interventions for autistic adults, no national clinical guidelines have been developed for psychological interventions aimed at this population in Denmark. Sparse evidence from profunctional studies conducted in England or Sweden is often referenced when attempting to guide psychological treatment approaches for this group [86]. To our knowledge, STEPS is the hitherto largest randomized clinical trial globally investigating the effect of VRSCT for autistic adults. The results of this innovative intervention approach may significantly advance research in the field of autism. If the VRSCT intervention in the STEPS trial proves effective, it has the potential to improve psychosocial functioning and quality of life, co-occurring clinical symptoms, and reduce social cognitive deficits in autistic adults. The intervention was co-developed in collaboration with autistic adults to ensure relevance and alignment with neuro-affirming principles. It is tailored to each participant’s goals, sensory needs, and communication preferences and aims to enhance comfort, confidence, and agency in social situations that participants themselves identify as meaningful. Rather than imposing neurotypical norms of interaction, the approach supports self-defined outcomes and promotes the transfer of skills into everyday life in ways that respect and value neurodivergent ways of relating. Establishing evidence-based interventions is essential to address the debilitating psychosocial challenges encountered by this population, especially considering the absence of established gold-standard interventions. Enhancing profunctional psychosocial skills can significantly impact the lives of these individuals by enabling them to build more meaningful and intimate relationships, better maintain personal and professional relations, and improve educational and occupational functioning. Developing effective interventions for autistic adults can further contribute to filling the critical gap in available support and thereby contribute to alleviating the personal and social consequences and the economic impact associated with autism. If proven effective, the manualized VRSCT intervention from this study could be implemented in mental health services across Denmark and potentially on an international scale, supported by training and ongoing supervision. Additionally, the intervention could be adapted to fit the needs of other clinical populations experiencing social cognitive deficits, such as autistic adolescents, individuals with bipolar disorder, or individuals with psychosis [87]. This could help address the increasing demand for novel, effective interventions for psychiatric conditions globally.

In this trial, eligibility was based on a clinical diagnosis of autism according to ICD-10 criteria. Although widely used in clinical and research settings, the diagnostic criteria in ICD-10 and DSM-5-TR have been criticized for overemphasizing observable behaviors and reflecting a prototypical male presentation of autism. These limitations can obscure how autism manifests differently across sex and age and may overlook how traits are masked by compensatory strategies, potentially contributing to the underdiagnosis of autism in females and the delayed identification of autism in adults [41,88]. While this may also affect the representativeness of the study sample by leading to fewer females being included, the final sex distribution will help determine the extent to which this concern is reflected in the sample.

The participants in the comparison group will receive TAU, administered across multiple sites. It is expected that there will be variations in the content and duration of TAU among participants. Although the specific details of TAU for each participant will be documented at the end of the trial, this variability remains a limitation. The inconsistency in TAU introduces less control compared to an active control uniformly administered in the same setting as the experimental intervention. However, TAU was chosen as the comparator because it reflects current clinical practice available for autistic adults in Denmark. Additionally, there is no established “gold standard” intervention specifically targeting social cognitive deficits in autistic adults, which otherwise could have been an obvious choice of comparator. Further, given the absence of a “gold standard” comparator and the choice of TAU as the control condition, it is important to note that the immersive and novel nature of the VR intervention may contribute to increased participant expectancy in the intervention group compared with TAU, which does not involve comparable technology. While a sham or low-dose VR control condition was considered to address this technology-expectancy bias, it was not feasible within the scope and resources of this study. To mitigate this risk, we included both a blinded, performance-based outcome measure (HiSoC) and an informant-rated version of the Social Responsiveness Scale (SRS-2AI) as secondary outcomes. This approach allows for triangulation across self-report, informant-report, and blinded assessment. Due to the exclusion criteria regarding autistic adults with medically untreated ADHD and autistic adults with intellectual disability, the results from this trial may not be generalizable to these populations. However, the findings may still provide useful indications of the potential effects for autistic adults with ADHD, and the intervention could eventually be adapted for autistic adults with intellectual disability. Finally, as the present trial does not include follow-up assessment beyond 6 months, the absence of longer-term follow-up (eg, at 12 months) limits conclusions regarding sustained effects. As the effects of VR-based interventions may diminish over time, and follow-up assessments beyond 3 months are often lacking [89], future trials should address this.

Autistic adults constitute a growing and largely overlooked population with limited evidence-based interventions available. To our knowledge, STEPS is the largest randomized clinical trial globally investigating the effect of VRSCT for autistic adults. If proven effective, VRSCT has the potential to improve psychosocial functioning and quality of life and reduce co-occurring clinical symptoms and social cognitive deficits in autistic adults, thereby potentially contributing to filling the critical gap in available support. The results of the trial should be interpreted in light of key limitations, including heterogeneity in TAU, potential technology-related expectancy effects, and follow-up limited to 6 months.

In light of these considerations, the findings of the trial are expected to advance research in the field of autism and may inform future clinical practice and subsequent trials, including in other clinical populations experiencing social cognitive challenges.