Discordant fetal growth, as defined by an intertwin weight difference (calculated as a percentage of the weight of the larger twin) of 20% or more, is associated with an increased perinatal mortality and morbidity. In this study, the accuracy of ultrasound estimation of fetal weight by abdominal circumference and biparietal diameter measurements in 107 twin pregnancies was assessed. Discordant growth greater than 20% as measured by ultrasound was found to have a sensitivity of 5-25%, a specificity of 87-95%, a positive predictive value of 12-43% and a negative predictive value of 80-85% at gestations ranging from 28 weeks to term. ROC calculations showed that changes in the discordance level used did not improve the value of the test. These results indicate that ultrasound measurement of fetal weight alone cannot be used accurately to assess discordant growth in twin pregnancies.