OBJECTIVES: To evaluate the performance of screening for small-for-gestational-age (SGA) fetuses by ultrasound biometry at 30-35 weeks' gestation, and to determine the impact of screening on obstetric and neonatal outcomes. METHODS: For this prospective cohort study, pregnant women were recruited from two French university maternity centers between 2003 and 2006. Performance measures of third-trimester biometry for the prediction of SGA, defined as estimated fetal weight < 10(th) centile, were analyzed. Obstetric outcomes and neonatal health status were compared, first, between SGA neonates diagnosed correctly at ultrasound examination (true positive (TP); n = 45) and SGA neonates that went undiagnosed (false negative (FN); n = 110) and, second, between non-SGA neonates identified as normal at ultrasound examination (true negative (TN); n = 1641) and non-SGA neonates diagnosed incorrectly as SGA (false positive (FP); n = 101). RESULTS: In the prediction of SGA, third-trimester ultrasound had a sensitivity of 29.0% (95% CI, 22.5-36.6%) and specificity of 94.2% (95% CI, 93.0-95.2%). Positive and negative predictive values were 30.8% (95% CI, 23.9-38.7%) and 93.7% (95% CI, 92.5-94.8%), respectively. One hundred and ten SGA neonates went undiagnosed at ultrasound. Compared to the TN neonates considered as of normal weight at ultrasound, planned preterm delivery (before 37 weeks) and elective Cesarean section for a fetal growth indication were 2.4 (P = 0.01) and 2.85 (P = 0.003) times more likely to occur, respectively, in the FP group of non-SGA neonates, diagnosed incorrectly as SGA during the antenatal period. There was no statistically significant difference in 5-min Apgar score < 7, cord blood pH at birth < 7.15 and need for neonatal resuscitation between the two subgroups (TN vs FP and TP vs FN). CONCLUSIONS: The performance of third-trimester ultrasound screening for SGA seems poor, as it misses the diagnosis of a large number of SGA neonates. The consequences of routine screening for SGA in a low-risk population may lead to unnecessary planned preterm deliveries and elective Cesarean sections in FP pregnancies, without improved neonatal outcome in the FN pregnancies.