Background Although exercise-based cardiac rehabilitation improves exercise capacity of coronary artery disease patients, it is unclear which training characteristic determines this improvement. Total energy expenditure and its constituent training characteristics (training intensity, session frequency, session duration and programme length) vary considerably among clinical trials, making it hard to compare studies directly. Therefore, we performed a systematic review and meta-regression analysis to assess the effect of total energy expenditure and its constituent training characteristics on exercise capacity. Methods We identified randomised controlled trials comparing continuous aerobic exercise training with usual care for patients with coronary artery disease. Studies were included when training intensity, session frequency, session duration and programme length was described, and exercise capacity was reported in peakVO2. Energy expenditure was calculated from the four training characteristics. The effect of training characteristics on exercise capacity was determined using mixed effects linear regression analyses. The analyses were performed with and without total energy expenditure as covariate. Results Twenty studies were included in the analyses. The mean difference in peakVO2 between the intervention group and control group was 3.97 ml·min− 1·kg− 1 (p < 0.01, 95% CI 2.86 to 5.07). Total energy expenditure was significantly related to improvement of exercise capacity (effect size 0.91 ml·min− 1·kg− 1 per 100 J·kg, p < 0.01, 95% CI 0.77 to 1.06), no effect was found for its constituent training characteristics after adjustment for total energy expenditure. Conclusions We conclude that the design of an exercise programme should primarily be aimed at optimising total energy expenditure rather than on one specific training characteristic.