Physical training improves glucose homeostasis in rats with experimental diabetes mellitus. The present study has been designed to ascertain if this beneficial effect is maintained or lost after a few days of inactivity. Male Wistar rats were injected with streptozotocin (STZ, 50 mg/kg) and those presenting one week later a blood glucose value between 250 and 400 mg/dl were retained in the protocol and randomly assigned to a sedentary (n = 18) or trained group (n = 23). An I.V. glucose tolerance test (0.5 g/kg) was performed in previously cannulated rats 64 h (trained rats; n = 15) or 12 days (detrained rats, n = 8) after the last bout of exercise. In comparison with results obtained in their sedentary counterparts (419 +/- 15 mg/dl), basal glucose levels were significantly lower in trained (362 +/- 20; p less than 0.05), but not in detrained rats (429 +/- 15; NS). Similar differences in plasma glucose levels were observed after glucose loading, even though the glucose disappearance rate constant was not significantly improved by training. Furthermore, basal insulin levels were significantly higher (p less than 0.01) in trained than in sedentary rats (20 +/- 3 vs 12 +/- 2 microU/ml) but such a difference had disappeared in detrained rats (9 +/- 2 microU/ml). These results indicate that the training-induced improvement in glucose homeostasis of diabetic rats is a transient phenomenon which is associated with an increase in circulating insulin levels. This suggests that the beneficial effect of training is not due solely to enhanced insulin sensitivity.