Virtual reality interventions for amblyopia: a systematic review with meta-analysis of visual outcomes

Amblyopia affects 2–5% of children and is the leading cause of monocular visual impairment in childhood. Traditional occlusion and atropine therapies are limited by poor adherence, particularly among school-age children. Virtual reality (VR)-based binocular interventions have emerged as an alternative approach, but evidence regarding their visual outcomes and clinical use remains heterogeneous. This systematic review evaluated visual outcomes and summarized treatment success, adherence, and adverse-event reporting associated with VR-based interventions for amblyopia, with particular attention to pediatric applicability. Six databases (PubMed, EMBASE, Scopus, Web of Science, CINAHL Ultimate, and CENTRAL) were searched for studies published between 1 January 2012 and 31 December 2025. Randomized and non-randomized quantitative studies evaluating VR-based interventions for amblyopia were eligible. Risk of bias and methodological quality were assessed using RoB 2, ROBINS-I, and design-specific JBI critical appraisal tools. Random-effects meta-analysis was undertaken only when outcomes, comparators, and measurement scales were sufficiently comparable. Nine studies were included. Comparator-specific BCVA findings were not pooled because comparator type and treatment exposure differed substantially. An exploratory within-group analysis showed a pooled BCVA change of − 0.14 logMAR (95% CI, − 0.21 to − 0.06; I² = 95%), and two studies reporting compatible log-transformed stereopsis outcomes showed a pooled change of − 0.33 log arcsec (95% CI, − 0.43 to − 0.22; I² = 0%). Treatment-success definitions differed and were not pooled. In a single mixed-age randomized trial with substantially different prescribed treatment burdens, adherence to at least 75% of prescribed treatment was higher with VR than with patching (OR = 23.22, 95% CI, 7.48–72.13). Adverse-event reporting was insufficiently standardized for quantitative synthesis. Seven of eight outcome-specific RoB 2 assessments were at high overall risk of bias, and the non-randomized comparative study was at serious overall risk according to ROBINS-I. Exploratory pre-post syntheses showed improvements in BCVA and stereopsis over time following VR-based interventions, but these non-comparative findings cannot establish treatment efficacy. Comparative evidence remains sparse and heterogeneous, while evidence for treatment success, adherence, and safety is limited. Larger pediatric randomized trials using standardized visual, adherence, and participant-level safety outcomes are needed. PROSPERO CRD420261411660.