OBJECTIVE: To test the hypotheses that, compared with participants using manual wheelchairs equipped with conventional rear anti-tip devices (C-RADs), those using a new RAD design that deploys through an arc (Arc-RAD) perform RAD-relevant wheelchair skills better and as safely. DESIGN: A randomized controlled study. SETTING: A rehabilitation center. PARTICIPANTS: Participants (N=30) including 16 able-bodied and 14 wheelchair users. INTERVENTION: Participants were provided with wheelchair skills training (up to 2.4h). MAIN OUTCOME MEASURES: Total percentage score on a set of 23 RAD-relevant skills of the Wheelchair Skills Test (WST, version 3.2) administered a minimum of 3 days after training. RESULTS: For the C-RAD and Arc-RAD groups, the mean +/- standard deviation RAD-relevant WST scores were 32.3%+/-8.5% and 85.1%+/-18.9% (Kruskal-Wallis, P<.001). Of the 23 RAD-relevant individual skills, the success rates for the Arc-RAD group were at least 20% higher (the criterion we set for clinical significance) in 17 (74%). For the C-RAD group, the success rate was 0% for the 12 wheelie-dependent skills, the 13-cm-high obstacle, and the 15-cm level change ascent. There were no serious adverse effects in either group. CONCLUSIONS: The new RAD design allows much better performance on relevant wheelchair skills than the conventional design without compromising safety.