We evaluated areas under receiver operating characteristic curves for sonographic estimated fetal weight (EFW) as a predictor of macrosomia in prolonged pregnancies. These areas were 0.85 for birth weights of > or = 4000 g and 0.95 for birth weights of > or = 4500 g. Both were significantly greater than 0.5, the area under curves for useless tests. Areas under curves before and after adjustment for time elapsed between measurement and delivery did not differ significantly. At the inflexion point cut-off level of 3711 g, sensitivity, specificity and positive and negative predictive values of EFW for birth weight of > or = 4000 g were 85, 72, 49 and 94%, respectively. At the inflexion point cut-off level of 4192 g for birth weight of > or = 4500 g, these values were 83, 92, 30 and 99%. The relative risk for birth weight of > or = 4000 g was 7.99, and for birth weight of > or = 4500 g, 39.50, both significant. In conclusion, sonographic EFW is a useful test for predicting macrosomia. Adjustment for time elapsed did not significantly improve either test, probably because of slow rates of fetal growth at this gestational age. Cut-off values derived from this analysis result in high sensitivity but low positive predictive value. A randomized controlled trial of mode and timing of delivery for predicted macrosomia is needed.