A prospective study was undertaken to test the hypothesis that insulin treatment in patients with gestational diabetes mellitus (GDM) with fasting plasma glucose (FPG) greater than 5.3 mM significantly reduces adverse perinatal outcome. Assigned to insulin or diet treatment based on FPG were 471 GDM women. Four factors believed to be associated with infants large for gestational age (LGA) were evaluated: FPG, overall glycemic control, maternal weight, and treatment regimen. We found that when glycemic control was optimized, the key factors related to large infants were FPG and treatment modality. In the low-FPG group (less than 5.3 mM), diet therapy achieved an incidence of 5.3% LGA. When insulin therapy was used to optimize control, an incidence of 3.5% LGA was found. Patients in the mid-FPG group (5.3-5.8 mM) had a higher increased rate of LGA (28.6%) for diet-treated versus insulin-treated women (10.3%). In addition, a fourfold increased risk for LGA was found in the diet-treated obese subjects in the mid-FPG group compared with insulin-treated obese women. Finally, treatment with insulin resulted in similar incidence of LGA within all FPG groups. We concluded that FPG greater than 5.3 mM can be the basis for initiation of insulin treatment in GDM subjects with optimization of glycemic control as the goal. This approach may contribute significantly to reduced neonatal risk and may foster a standardized method for rapid and effective assignment to treatment.