OBJECTIVE: Assess and compare accuracy of birth weight prediction using a combination of two-dimensional (abdominal circumference - AC and femur length - FL) and three-dimensional parameters (fetal arm -VolArm and thigh -VolTh volumes). METHODS: A cross sectional study was carried out involving 78 singleton, live, euploid fetuses without structural malformations born within 48 hours after ultrasonography. VolArm and VolTh were obtained by three-dimensional ultrasound using the multiplanar mode at 5 mm intervals. AC and FL were measured by two-dimensional ultrasound. Linear and polynomial regressions were calculated to determine the best formula to predict birth weight using VolArm, VolTh , CA and FL. ANOVA was used to compare errors in birth weight prediction using this formula and those obtained using the Shepard and Hadlock formulae. RESULTS: The best formula for prediction of birth weight was a simple linear regression (Weight = -1486.1 + 60.5AC + 140.57FL + 16.6VolArm + 4.8VolTh), R2= 0.932. The error (E), absolute error (AE), percent error (PE) and absolute percent error (APE) for this new formula were 0 g, 0.2%, 112.2 g and 3.7%. This new formula had smaller E, AE, PE and APE than the Shepard formula and smaller AE and APE than Hadlock´s formula. CONCLUSION: A formula using VolArm, VolTh, AC and FL was more accurate for prediction of birth weight than formulae using only two-dimensional parameters.
Fetal weight; Imaging; Three-Dimensional; Ultrasonography
