Introduction To evaluate the accuracy and validity of common formulae for estimating fetal weights using ultrasonographic measurements in low birthweight babies (LBW) ≤2500 g and high birth weight (HBW) ≥4000 g.

Methods All births within the study groups over a one year period were identified. Maternal demographic data and most recent ultrasonographic data from fetal well-being scans within 10 days of delivery were collected. Estimated fetal weight was calculated using biometric data and compared to actual birth weights, giving mean percentage error and standard deviation for each formulae within the groups.

Results All formulae show systematic errors across both groups, with a general tendency to overestimate weight of LBW infants and underestimate the weight of those within the HBW group. In the LBW group, the Hadlock AC-FL method had the smallest systematic error (95% CI: 7.2%±2.5%). In the HBW group, the Woo BPD-AC-FL formula showed the lowest systematic error (95% CI: -4.9%±2.9%). The Hadlock AC-FL method again performed well (95% CI: -5.7%±2.0%).

Conclusions Although current reference ranges are based on Caucasian populations the error in estimating fetal weight in a local population of mainly non-Caucasian ethnicity, is still within an acceptable margin of error (less than 10%). Overall, methods which are not based upon biometric measurements of the fetal head appeared to be more accurate. Customised growth charts may improve accuracy by taking into account non-pathological variants of fetal growth (ethnicity and parity and maternal weight).