When extremely preterm birth is anticipated, a reliable estimate of neonatal outcome is essential for the parents and health care providers who face difficult management decisions. Estimates of birth weight and gestational age are most commonly used for this purpose. The National Institute of Child Health and Human Development Maternal-Fetal Medicine Units (NICHD MFMU) Network performed an observational study of data available before delivery of infants with birth weights < 1000 g. Ultrasonographic variables (estimated fetal weight, obstetrically estimated gestational age, femur length, and biparietal diameter) and clinical variables (maternal race, antenatal care, substance abuse, medical treatment, reason for delivery, fetal gender, and presentation) were studied as predictors of intrapartum stillbirth, neonatal death, survival to 120 days after birth or to discharge, and with markers of "serious" morbidity (high-grade intraventricular hemorrhage, retinopathy of prematurity, necrotizing enterocolitis, oxygen dependence at discharge or 120 days, and seizures). Survival without serious morbidity was considered "intact." Logistic regression was used to evaluate the influence of the obstetrician's opinion of viability and willingness to perform cesarean delivery for fetal distress, birth weight, growth, gender, presentation, and ethnicity on outcomes. Fetal femur length and estimated gestational age predicted survival better than did biparietal diameter or estimated fetal weight. Antenatal ultrasound and clinical data did not distinguish those infants who would suffer serious morbidity or be considered intact. Willingness to perform cesarean delivery was associated with increased likelihood of both survival and intact survival by virtually eliminating intrapartum stillbirth and reducing neonatal mortality. However, such practice was associated with an increased chance of serious morbidity among survivors below 800 g or 26 weeks'. Although obstetricians were willing to intervene for fetal indications in most cases by 24 weeks', willingness to perform cesarean delivery was associated with twice the risk for serious morbidity in survivors at that gestational age.