Purpose: (1) to analyze training characteristics of recreationally active wheelchair users during handcycle training, and (2) to examine the associations between training load and change in physical capacity. Methods: Former rehabilitation patients (N = 60) with health conditions such as spinal cord injury or amputation were included. Participants trained for five months. A handcycling/arm crank graded exercise test was performed before and after the training period. Outcomes: peak power output per kg (POpeak/kg) and peak oxygen uptake per kg (VO 2peak/kg). Training load was defined as Training Impulse (TRIMP), which is rating of perceived exertion (sRPE) multiplied by duration of the session, in arbitrary units (AU). Training intensity distribution (TID) was also determined (time in zone 1, RPE ≤4; zone 2, RPE 5–6; zone 3, RPE ≥7). Results: Multilevel regression analyses showed that TRIMP sRPE was not significantly associated with change in physical capacity. Time in zone 2 (RPE 5–6) was significantly associated with ΔVO 2peak, %ΔVO 2peak, ΔVO 2peak/kg and %ΔVO 2peak/kg. Conclusion: Training at RPE 5–6 was the only determinant that was significantly associated with improvement in physical capacity. Additional controlled studies are necessary to demonstrate causality and gather more information about its usefulness, and optimal handcycle training regimes for recreationally active wheelchair users.IMPLICATIONS FOR REHABILITATION Monitoring of handcycle training load is important to structure the training effort and intensity over time and to eventually optimize performance capacity. This is especially important for relatively untrained wheelchair users, who have a low physical capacity and a high risk of overuse injuries and shoulder pain. Training load can be easily calculated by multiplying the intensity of the training (RPE 0–10) with the duration of the training in minutes. Results on handcycle training at RPE 5–6 intensity in recreationally active wheelchair users suggests to be promising and should be further investigated with controlled studies.