One hundred eight gestational diabetics were randomized to receive either diet alone or diet plus insulin (20 units NPH and 10 units regular) for glycemic control. Blood glucose levels were evaluated weekly in a high-risk clinic where medical and nutritional support and counseling were provided. Among 68 women successfully treated for a minimum of 6 weeks, the mean birth weight, macrosomia rate, and ponderal index were reduced significantly in the insulin-treated group. Insulin reduced birth weights significantly in women with a delivery weight of 200 lb or more (4060 +/- 342 versus 3397 +/- 640 g) and in those with a delivery weight less than 200 lb (3324 +/- 448 versus 3047 +/- 394 g). No patient with good glucose control and a maternal delivery weight under 200 lb had a newborn over 4000 g. Patients failing glycemic control were at greatest risk (30%) for fetal overgrowth whether initially receiving insulin or not. Maternal obesity or failure to achieve glycemic control should alert the clinician to a substantially increased risk of macrosomia.