Estimation of birth weight by two-dimensional ultrasonography: a critical appraisal of its accuracy.

OBJECTIVE: To assess the accuracy and characterize two-dimensional ultrasonographic formulas for the estimation of birth weight according to the type of fetal biometric parameters these formulas rely on to make fetal weight predictions. METHODS: A prospective recruitment of 589 pregnant women was carried out for this cross-sectional study. Different biometric parameters were taken ultrasonographically to estimate birth weight using 35 different formulas. Only those patients who delivered within 48 hours were considered for the analysis (n=441). Differences between the estimated and actual birth weight were assessed by percentage error, accuracy in predictions within +/-10% and +/-15% of error, and use of the Bland-Altman method. All formulas were assessed individually and clustered on the basis of the type of fetal biometric information that they incorporate. RESULTS: Twenty-nine formulas provided an overall mean absolute percentage error less than or equal to 10%, with overall predictions within +/-10% and +/-15% of the actual birth weight (69.2% and 86.5%, respectively). Twenty formulas showed a good accuracy (bias 0.50 or less) and low variability (mean standard deviation 1.2). Among the categorized algorithms, formulas based on head-abdomen-femur measurements showed the lowest mean absolute percentage error. Upon stratification for birth weight, the group of formulas that rely on abdomen and femur measurements performed best for fetuses weighing more than 3,500 g (P<.01). CONCLUSION: Our findings show that most formulas are relatively accurate at predicting birth weight up to 3,500 g, and all algorithms tend to underestimate large fetuses. LEVEL OF EVIDENCE: III.