Virtual, augmented, and mixed reality training for improving social skills in individuals with Autism Spectrum Disorder: A systematic review.

All seven studies were rated at unclear risk of bias overall, so no result rests on a study the reviewers judged clean. The review's own reporting of risk of bias is contradictory: one statement says all studies were unclear on the randomisation process, while another says all but one were at low risk. Most studies were small. Three had between 14 and 43 participants, which the authors say may have limited the precision of the estimates. The studies differed so much in the type of training, the ages of participants and the way social skills were measured that the authors say the relative benefits of the different approaches remain unclear, and they could not identify how often training should be given. No meta-analysis was performed, so there is no pooled estimate of effect. Only published articles found in the search were included; unpublished data and review articles were excluded, and no journals were hand-searched. Some effect estimates had ranges that crossed zero, so the improvement was not clearly demonstrated. One included trial gave VR alongside risperidone, so the drug's contribution cannot be separated from the VR's, and the review reported no effect size for that trial.

The authors received no financial support for the research, authorship or publication of this article, and declared no conflicts of interest.

Do not read this as proof that virtual or mixed reality training improves social skills in autism. The review's own conclusion is that the supporting evidence is limited, the studies were too different to compare, no meta-analysis was done, and every study was at unclear risk of bias. One trial gave VR alongside risperidone and another gave VR-CBT alongside a learning style programme, so no single ingredient can be credited from those results.