Introduction It is still unknown how best to maintain effects of exercise programmes in COPD in the long-term. We present the long-term effects of a community-based exercise programme incorporated in a self-management programme, compared to a self-management programme only in patients with COPD. Methods All included patients participated in four self-management sessions. Additionally, patients in the intervention group participated in an 11-month community-based exercise programme led by physiotherapists. Patients trained three times/week for six months and two times/week during the subsequent five months. To encourage a behavioural change towards exercise, one of these weekly training sessions was home-based (unsupervised). No formal exercise training was offered to intervention patients in the second year. Results The intervention was assigned to 80 patients, and the control condition to 79 patients. 82.5% and 78.5% of the intervention and control group, respectively, completed 24 months follow-up. Modified intention-to-treat analyses were performed. Although statistically significant after 12 months (35.1 m (95%CI: 8.4–61.8)), the between-group difference on maximal exercise capacity was not statistically significant after 24 months (12.2 m (95%CI: −16.6 to 41.0). Nevertheless, the between-group difference in daily physical activity was maintained after 24 months (1193 steps/day (95%CI: 203–2182)). A beneficial effect was also found on CRQ dyspnoea score but not on other CRQ domains, CCQ and HADS. Conclusions Our intervention was effective in achieving a behavioural change reflected by a sustained increase in daily physical activity, not accompanied by a sustained increase in maximal exercise capacity after two years of follow-up (ISRCTN81447311).