Selection of standards for estimated fetal weight percentiles

To evaluate which of 10 standards or references should be used for determination of percentiles of estimated fetal weight (EFW).

Cross-sectional study using pooled data from four maternal-fetal medicine (MFM) practices. Inclusion criteria were as follows: ultrasound exam during 2022, singleton fetus, gestational age 24^0/7 to 39^6/7 weeks, biometry measured, and fetal cardiac activity present. EFW was calculated using the three-parameter formula of Hadlock et al. We evaluated five American birth weight (BW) references and five fetal weight (FW) standards. The fit of each reference or standard to our EFW distribution was assessed by seven criteria: mean z-score close to 0, standard deviation (SD) of z close to 1, Kolmogorov-Smirnov D-statistic close to 0, close to 10% of EFWs <10th percentile, close to 10% of EFWs > 90th percentile, and Youden J-statistics close to 1 at the 10th and 90th percentiles.

In 20,643 ultrasound exams, the formulas from the 1991 FW standard of Hadlock et al. had the best fit to our distribution of EFW (mean z-score 0.07, SD of z 1.04, D-statistic 0.036, J-statistics 0.946 and 0.972 at 10th and 90th percentile, respectively). All five EFW standards had better fit than any of the BW references, primarily because the range of normal of the BW references was much larger than the normal range of observed EFWs, resulting in underdiagnosis of abnormal EFW (both EFW <10th percentile and EFW >90th percentile). The range of normal from the Hadlock table was larger than that from the Hadlock formulas and the observed range, resulting in fewer diagnoses of abnormal EFW. In the subgroup of patients at low risk for small for gestational age or large for gestational age, the Hadlock formulas still had the best fit to our EFW distribution.

Of the norms studied, the FW distribution from the Hadlock 1991 formulas had the best fit to our observed EFW distribution and are the most appropriate standard for determination of EFW percentiles in our US-based MFM practices. All the FW standards had better fit to our data than any of the BW references, so EFW percentiles should be based on FW standards, especially the Hadlock standard, not BW references.