This study was designed to ascertain if the improvement in glucose homeostasis found in diabetic animals submitted to physical training is due to an increased secretion of epinephrine by the adrenal medullae. Male Wistar rats were surgically adrenodemedullated (ADM group) or sham-operated (SHAM group). After a 3-week recovery period, a bolus of streptozotocin (40 mg/kg) was injected intravenously (IV), and the animals presenting 1 week later with a blood glucose value between 14 and 22 mmol/L were retained in the protocol and randomly assigned to a sedentary (SHAM-DS and ADM-DS) or trained (SHAM-DT and ADM-DT) group. Physical training was done on a treadmill according to a 10-week progressive program. An IV glucose tolerance test (0.5 g/kg) was performed in previously cannulated rats, 64 hours after the last bout of exercise. Pancreatic insulin and glucagon content was also determined. In sedentary diabetic rats, adrenodemedullation had no effect on plasma glucose, insulin, or glucagon levels, neither in the basal state nor following the glucose load. Basal glucose levels were diminished by training in both SHAM (16.1 +/- 1.5 v 21.8 +/- 0.4 mmol/L; P less than .01) and ADM (12.4 +/- 1.7 v 21.1 +/- 1.2 mmol/L; P less than .01) groups, with values lower in ADM-DT than in SHAM-DT rats (P less than .05). After glucose loading, the glucose levels were significantly lower (P less than .01) throughout the test in both SHAM-DT and ADM-DT rats than in their sedentary counterparts.(ABSTRACT TRUNCATED AT 250 WORDS)