This study presents a mathematical model by which power output (PO) delivered to the rear wheel during handrim wheelchair propulsion on a roller ergometer can be determined for individual wheelchair-user combinations. PO is calculated from the torque applied to the wheel and its angular velocity. The torque applied is a function of one total internal torque of the wheelchair-ergometer system, the rotational moment of inertia of the rear wheel, the one of the roller and its angular acceleration. The total internal torque reflects all internal friction forces and is determined with a deceleration test. To assess the reliability of this approach, 11 able-bodied subjects underwent progressively increasing exercise tests on two different occasions. PO values ranged from 12 to 63 W and were highly reliable (r2 > 0.95). Peak physiological responses were never different from test 1 and 2 (repeated measures ANOVA; p: N.S.) and correlations were 0.90, 0.72, 0.88, 0.82, 0.70 and 0.85 for PO, oxygen uptake (VO2), heart rate, minute ventilation, carbon dioxide production and blood lactate concentration, respectively. After an initial increase, gross mechanical efficiency dropped at higher velocities, with values ranging from 4.64 to 11.26%. In conclusion, the roller ergometer, the mathematical model to determine PO and the protocol used seem to be adequate to exercise test people in a handrim wheelchair. It is feasible to apply the theoretical procedure to other roller ergometers which would allow for comparisons of exercise intensities and protocols between different devices used in exercise physiology and rehabilitation.