A modified oral glucose tolerance test was done during the third trimester in 811 pregnancies in Pima Indian women over a 13-yr period, and maternal and fetal complications were documented. Diabetes was known to be present in 51 pregnancies. Among those who were not previously known to have diabetes, rates of perinatal mortality, macrosomia, toxemia, and cesarean section varied directly with glucose concentration, but congenital malformation and prematurity rates did not. Rates of all of these complications were higher in known diabetic women than in the remainder of the population. In addition to glucose concentrations, maternal weight and age were predictive of macrosomia and toxemia. Third-trimester glucosuria was found to be of very limited value as a screening procedure for gestational diabetes. In 233 women followed for 4-8 yr, the third-trimester glucose concentration was highly predictive of the subsequent incidence of diabetes.