Effectiveness of VR-based cognitive training and games on cognitive rehabilitation in patients with MCI: a systematic review and meta-analysis.

Background: Mild Cognitive Impairment (MCI) represents a prodromal dementia stage marked by cognitive decline without functional impairment. Given limited drug efficacy and global aging, non-pharmacological interventions are urgently needed. Virtual reality (VR) enables immersive cognitive rehabilitation, yet evidence remains inconsistent due to divergent intervention approaches (training vs. gaming) and technical parameters like immersion level. Objective: This systematic review and meta-analysis synthesized evidence from randomized controlled trials (RCTs) to evaluate the efficacy of VR-based cognitive training and gaming interventions on cognitive function in older adults with MCI and to investigate the moderating role of immersion level. Methods: We systematically searched four electronic databases (PubMed, Web of Science, Embase, Scopus) from inception to July 20, 2025, for RCTs investigating VR interventions (cognitive training or games) in individuals aged ≥ 55 years diagnosed with MCI. Two independent reviewers performed study selection, data extraction (including intervention characteristics, implementation details, and behavior change techniques), and risk-of-bias assessment using the Cochrane Risk of Bias tool (RevMan 5.4.1). Standardized mean differences (Hedges's g) with 95% confidence intervals (CI) were pooled using random-effects models in Stata 18.0. Heterogeneity was quantified using I2 . Publication bias was assessed via funnel plots and Egger's test. Pre-specified meta-regression explored immersion level as a potential moderator. The certainty of evidence was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system. Results: Of the 2,486 articles retrieved in total, 11 studies were included in the analysis. VR demonstrated a statistically significant improvement in the efficacy of cognitive rehabilitation among patients with MCI (Hedges's g = 0.6, 95% CI: 0.29 to 0.90, p < 0.05). Specifically, VR-based games (Hedges's g = 0.68, 95% CI: 0.12 to 1.24, p = 0.02) showed greater advantages in improving cognitive impairments compared to VR-based cognitive training (Hedges's g = 0.52, 95% CI: 0.15 to 0.89, p = 0.05). The immersive level of VR interventions emerged as a significant moderator of heterogeneity across the included studies. Based on the GRADE criteria, the quality of evidence for the efficacy of VR-based interventions on cognitive function in individuals with MCI is moderate. A stratified analysis by intervention type showed that VR cognitive training is supported by moderate-certainty evidence, while evidence for VR games is of low certainty. Conclusion: VR-based interventions, including cognitive training and games, effectively improve cognitive function in MCI patients, with VR games showing a trend toward greater efficacy. Immersion level critically influences therapeutic outcomes, requiring optimized sensory integration while accommodating individual tolerance. These findings support supervised clinical VR training alongside engaging home-based protocols to enhance adherence. Future development of standardized immersion adjustment and personalized guidelines will advance utility across care settings.