The estimation of birth weight is an important aspect of decisions about the management of late pregnancy and delivery. Experienced clinicians are able to estimate birth weight within 500 g (1 lb, 1 oz) in 85 percent of average-weight term infants, but this percentage decreases for preterm or macrosomic fetuses. The average error in birth weight predicted by sonogram is estimated to be between 6 and 15 percent. Again, the average error increases in infants weighing less than 2,500 g (5 lb, 8 oz) or more than 4,000 g (8 lb, 11 oz). Several strategies have been advocated to provide the best prediction of birth weight, especially in tiny or unusually large fetuses. Multiple ultrasonographic examinations in late pregnancy have been reported to improve predictions but have significant costs. Pressman and colleagues compared the predictive ability of sonographic examinations early in the third trimester with examinations conducted after 37 weeks' gestation.

They studied healthy singleton pregnancies booked for delivery at a university hospital. Each patient underwent sonography at least once before 37 weeks' and again after 37 weeks' gestation. All sonographic examinations were carried out by the same team using the same techniques of examination and calculation of anticipated birth weight. Data were collected on the mother, pregnancy and infant (following delivery).

The 138 mothers underwent 276 ultrasonographic examinations. In 22 cases, no indication was provided for the two third-trimester sonographic examinations. The other cases included one or more indications, such as risk for macrosomia (30 mothers), risk for growth restriction (61 mothers) and suspected fetal abnormality (22 cases). Sonograms obtained before 37 weeks resulted in fewer errors in predicting true birth weight than sonograms obtained after 37 weeks' gestation. This finding was demonstrated for all deliveries and when cases of suspected growth abnormalities were studied independently. The differences were slight (less than 100 g [3.5 oz]) and probably were statistically but not clinically significant.

The authors conclude that a single sonogram obtained between 34 and 36.9 weeks' gestation provided comparable and marginally more accurate prediction of birth weight for all pregnancies, including those of suspected growth abnormality, than sonograms obtained closer to term. They recommend that the strategy of a single sonogram obtained between 34 and 37 weeks' gestation replace serial sonography or sonograms taken late in pregnancy.