The cause of macrosomia in the infant of the diabetic woman is still not completely defined. The National Institute of Child Health and Human Development--Diabetes in Early Pregnancy Study, which recruited insulin-dependent diabetic and control women before conception, provided an opportunity to address the relationship between maternal glycemia and percentile birth weight. Data were analyzed from 323 diabetic and 361 control women. Fasting and nonfasting venous plasma glucose were measured on alternate weeks in the first trimester and monthly thereafter. Glycosylated hemoglobin was measured weekly in the first trimester and monthly thereafter. More infants of the diabetic women were at or above the 90th percentile for birth weight than infants of control women (28.5% versus 13.1%, p less than 0.001). Although first-trimester nonfasting glucose and glycosylated hemoglobin levels were positively correlated with infant birth weight (p less than 0.001 and p = 0.008), when the analyses were adjusted for the variables of the subsequent trimesters the values became insignificant, whereas the third-trimester nonfasting glucose levels adjusted for values in prior trimesters emerged as the stronger predictor of percentile birth weight (p = 0.001). After adjusting for maternal hypertension, smoking, and ponderal index, the above relationships remained. In conclusion, monitoring of nonfasting glucose levels rather than the fasting levels, which are more commonly monitored in clinical practice, are necessary to prevent macrosomia.