This prospective controlled trial included two groups of subjects with coronary artery disease: one of 33 patients with non-insulin-dependent diabetes mellitus (6 females and 27 males, mean age 57 +/- 9 years, left ventricular ejection fraction of 56 +/- 14%) and the control group of 33 subjects without diabetes (3 females and 33 males, mean age of 57 +/- 11 years, LVEF of 58 +/- 11%). Before training heart rate was faster for patients with diabetes at rest (72 +/- 11 vs 66 +/- 81 bpm, p = 0.03) and at the end of the stress test (127 +/- 15 versus 118 +/- 18 bpm, p = 0.03). After rehabilitation, resting heart rate and peak heart rate were similar. However cardiovascular capacities improvement was better in subjects without diabetes mellitus, especially concerning peak VO2 (28.98 +/- 8.88 versus 22.78 +/- 6.28 mL/min/kg, p < 0.01) and mechanical power (138 +/- 48 versus 118 +/- 23 watts, p < 0.01). Two groups were retrospectively distinguished among diabetic patients: one group of 17 patients showing a VO2 improvement superior to 5% and a second group non improved (VO2 < 5%). For the second group the two selective factors were higher fasting glycemia (1.83 +/- 0.75 versus 1.31 +/- 0.38 g/L, p = 0.01) and higher hemoglobin A1C (8.05 +/- 2.04 vs 6.62 +/- 1.03%, p = 0.02). Heart rate variability was not significantly different for these two groups: changes in autonomic nervous system cannot explain resistance of diabetic subjects to training. On the other hand, principal aim must be the equilibrium of glycemia in the management of diabetes.