We synthesized studies published since 2000 that assessed the effects of using virtual manipulatives to increase the mathematical accuracy of students with disabilities. We extracted a total of 1,796 raw data points from 114 cases across 35 single-case studies. By applying three-level multilevel modeling, we analyzed both immediate effects and trends during the intervention phase as well as moderation effects related to student characteristics (case level) and intervention features (study level). Both the average immediate effect and trend during the intervention were statistically significant. The average immediate effect varied significantly by student grade, disability type, developer, device, type of virtual manipulative, and visual model embedded in virtual manipulatives. Neither student characteristics nor intervention feature-related moderators significantly influenced the average trend during the use of virtual manipulatives.