To evaluate the accuracy of differentul trasonographic fetal weight estimating formulas in gestational diabetes mellitus.

A retrospective analysis of 92 singleton gestational diabetes maternal births from January 2016 to June 2017 was performed at Huzhou Maternal and Child Health Hospital. The mean percent error (MPE) and mean percent absolute error (MAPE) between the estimated fetal body mass (EFW) and the actual postnatal mass (ABW) assessed by different ultrasound formulas were compared. The intra-group correlation coefficient (ICC) was used to evaluate the consistency between EFW and ABW assessed by different ultrasound formulas. The Kruskal-Wallis rank sum test was used to compare the MPE and MAPE of different fetal formulas to evaluate the birth weight of the fetus, and the different ultrasound formulas were used to evaluate the MAPE of the birth weight of different body weights. The Mann-Whitney test was used to compare the two pairs. The χ² test was used to compare the accuracy of 8 ultrasound formulas evaluating the birth weight of the 4000 g of the fetus (overestimation rate, underestimation rate) and the degree of bias (MAPE≤5%, ≤10% and ≤15%), and the χ² test was used for further comparison.

The ICCs of the 8 commonly used ultrasound formulas were quite different, of which the formula 1 (Hadlock) and formula 7 (Schild) had the highest ICC (0.907 and 0.925) in all the 8 formulae. There was a significant difference between the average percent error in weight assessment and the absolute mean absolute error. The MPE and MAPE of Formula 1 (Hadlock), Formula 2 (Hadlock) and Formula 7 (Schild) were significantly lower than those of other formulas. All the 8 formulas had statistically significant differences for the fetuses>4000 g. The MAPEs of formula 1 (Hadlock) and formula 7 (Schild) were significantly lower than those of other formulas, while that of the formula 7 was significantly lower than that of formula 1. There were significant differences in the overestimation rate, underestimation rate, and bias≤10% for the>4000 g fetus of different formulas. Overestimated and underestimated rates of formula 1 (Hadlock) and formula 7 (Schild) for fetus>4000 g were significantly lower than those of other formulas. The ratio of formula 1 and formula 7 with bias≤10% was significantly higher than those of other formulas.

In the commonly used eight formulas, two formulas proposed by Hadlock and Schild have higher efficacy in the gestational diabetes mellitus and can reduce the possibility of over estimating or underestimating.