We studied the impact of a training program on glucose tolerance in gestational diabetes mellitus. Women with gestational diabetes mellitus (N = 19) were randomized into either group I, a 6-week diet alone group (24 to 30 kcal/kg/24 hours; 20% protein, 40% carbohydrate, 40% fat), or group II, which followed the same diet plus exercise (20 minutes three times a week for 6 weeks). An arm ergometer was used to maintain heart rate in the training range. Glycemic response was monitored by glycosylated hemoglobin, a 50 gm oral glucose challenge with a fasting and 1-hour plasma glucose, and blood glucose self-monitoring, fasting and 1 hour after meals. Week 1 glycemic parameters were the same for both groups. Week 6 data (mean +/- SD) were as follows: group I glycosylated hemoglobin, 4.7% + 0.2% versus group II, 4.2% +/- 0.2%; p less than 0.001. The group I glucose challenge fasting value was 87.6 +/- 6.2 versus 70.1 +/- 6.6 mg/dl, p less than 0.001 for group II. The group I 1-hour plasma glucose challenge result was 187.5 +/- 12.9 mg/dl versus 105.9 +/- 18.9 mg/dl for group II, p less than 0.001. The glycemic levels diverged between the groups at week 4. We conclude that arm ergometer training is feasible in women with gestational diabetes mellitus and results in lower glycosylated hemoglobin, fasting, and 1-hour plasma glucose concentrations than diet alone. Arm ergometer training may provide a useful treatment option for women with gestational diabetes mellitus and may obviate insulin treatment.