Our aim was to evaluate the ability of prenatal ultrasound scans to predict fetal growth discordance, and to examine the correlation between fetal weight estimated by ultrasound with actual birthweight in twin pregnancies. The study consisted of 221 twin pregnancies with ultrasound fetal weight estimates based on Hadlock's 4 parameter formula. Prediction of intertwin birthweight discordance was examined at 4 different intervals between ultrasound examination and delivery (0–7 days, n = 96; 8–14 days, n = 66; 15–21 days, n = 58; 22–28 days, n = 59 pregnancies), with a total of 279 ultrasound examinations. Birthweight discordance was considered as a difference of 20% or greater. The correlation between fetal weight estimated between 0 and 7 days and actual birthweight was calculated by intraclass correlation coefficient. The predictive values for intertwin discordance of 20% or more in the 0 to 7 days group were: sensitivity = 93.6%, specificity = 79.4%, positive predictive value = 89.2%, negative predictive value = 87.1% and accuracy = 88.6%. In the groups with scans carried out between 8 and 14 days, 15 and 21 days, and 22 and 28 days, the sensitivity and accuracy values were 95.8% and 84.9%, 95.6% and 84.5%, 90.9% and 84.8%, respectively. Fetal growth discordance in twins can be accurately predicted by ultrasound examination performed up to 28 days before birth. There is a good correlation between fetal weight estimated between 0 and 7 days and actual birthweight.