Estimating birth weight in twins. Comparison of eight sonographic models

To establish the relative accuracy of eight sonographic models that use one to four fetal biometric measurements to derive birth weight in twin pregnancies.

Over three years, 181 twin gestations with sonographic mensuration of biparietal diameter, head circumference, abdominal circumference (AC) and femur length (FL) within 96 hours of delivery were analyzed. Analysis of variance followed by Student-Newman-Keul's test was used to determine the relative accuracy of the eight models. A receiver-operating characteristic (ROC) curve was constructed to determine the sonographic estimate that would consistently identify a nonvertex twin B with a birth weight of < or = 1,500 g.

The mean gestational age was 32.4 +/- 4.3 (SD) weeks, and the mean birth weights for twins A and B were 1,851 +/- 676 and 1,883 +/- 655 g, respectively. For both fetuses, the mean standardized absolute error was significantly higher for estimates based on one fetal parameter (AC or FL). Regardless of presentation (A or B, vertex or nonvertex) or whether the actual birth weight was < or = 1,500 or > 1,500 g, accuracy was similar between the six models that used two or more fetal parameters to predict the newborn's weight. ROC analysis revealed that among nonvertex twins B, an estimate of birth weight (based on BPD and AC or AC and FL) < or = 1,700 g will consistently identify the newborn with a birth weight < or = 1,500 g.

Sonographic models that use two fetal parameters are adequate for assessing birth weight in twin gestations regardless of position or if the birth weight is < or = or > 1,500 g.