The purpose of this study was to compare the cardiorespiratory response and mechanical efficiency (ME) of highly trained spinal cord injured (SCI) handcyclists with untrained SCI men. Ten trained handcyclists (=2 years training) and ten untrained but physically active SCI men completed an incremental exercise test to exhaustion and a sub-maximal test (50 and 80 W) on an electromagnetically braked arm ergometer. The trained participants completed a questionnaire on their training and race performance over the past year, including average training volume (in kilometers), number of training sessions per week and best 20-km time trial. The trained SCI men had higher [(V)/dot] /textO/user22 Unknown control sequence '/user'peak, peak power (p = 0.001) and peak heart rate (p = 0.021) compared to the untrained SCI men. The trained men had higher (p = 0.001) ME at 50 W (14.1 ᠲ.0%) and 80 W (17.2 ᠲ.6) compared to the untrained men (50 W; 12.5 ᠱ.8 and 80 W; 15.7 ᠲ.1). Peak power (r = -0.87, p = 0.001), [(V)/dot] /textO/user22 Unknown control sequence '/user'peak (r = -0.67, p = 0.033) and ME (r = -0.58, p = 0.041) were negatively correlated with the participants best 20-km time trial. Multiple linear regression indicated peak power (p < 0.001) and [(V)/dot] /textO/user22 Unknown control sequence '/user'peak (p = 0.021) were the best predictors (87%) of 20-km time trial performance. Highly trained SCI handcyclists have a greater aerobic capacity and ME compared to untrained SCI, and are able to reach their maximum age-predicted heart rate during an incremental exercise test. The best predictor of 20 km race performance in highly trained SCI handcyclists is peak power attained during an incremental exercise test.