Receiver operator characteristic curves for both clinical and sonographic predictions of macrosomia subsume areas between 0.81 and 0.95, significantly larger than the area of 0.5 that indicates a useless test. Thus, these tests are defined as useful from a statistical point of view. Prediction of macrosomia by clinical or imaging techniques, however, is limited by the substantial false-positive and false-negative rates inherent in these tests. We recommend that physicians continue to use clinical methods to estimate fetal weight, including asking women with parity to provide their own estimates. We recognize that the relative error associated with clinical or sonographic estimates of fetal weight limits their use in clinical practice. Sonographic laboratories may improve their results by performing ROC curve analysis on their own data and by selecting cutoff values that best predict macrosomia in their setting. Serial sonographic measurements that are above the limits chosen to define macrosomia increase the likelihood that a birth weight will be macrosomic. Separate ROC curves must be generated for twins and breech presentations and for patients with diabetes to answer weight-related clinical questions such as mode and timing of delivery. Three-dimensional ultrasound and magnetic resonance imaging are expected to generate ROC curves for estimates of fetal weight that are better than those for two-dimensional ultrasound or clinical estimates. Such analyses have yet to be published.