Department of Obstetrics and Gynecology, Center for Integral Assistance to Women's Health, State University of Campinas Medical School, Campinas, Brazil.
Ultrasound Obstet Gynecol. 2010 Apr;35(4):426-33. doi: 10.1002/uog.7518.
To compare the accuracies of birth-weight predicting models derived from two-dimensional (2D) ultrasound parameters and from total fetal thigh volumes measured by three-dimensional (3D) ultrasound imaging; and to compare the performances of these formulae with those of previously published equations.
A total of 210 patients were evaluated to create a formula-generating group (n = 150) and a prospective-validation group (n = 60). Polynomial regression analysis was performed on the first group to generate one equation based on 2D ultrasound measurements, one based on fetal thigh volume measured by the multiplanar technique (ThiM) and one based on fetal thigh volume obtained by the Virtual Organ Computer-aided AnaLysis (VOCAL()) method (ThiV). Paired-samples t-tests with Bonferroni adjustments were used to compare the performances of these equations in the formula-finding and the prospective-validation groups. The same approach was used to compare the accuracies of the new 2D and 3D formulae with those of both original and modified 2D equations from previous publications, as well as the 3D model reported by Chang et al.
The formulae with the best fit for the prediction of birth weight were: estimated fetal weight (EFW) = - 562.824 + 11.962x AC x FDL + 0.009 x BPD(2)x AC(2) (where AC is abdominal circumference, FDL is femur diaphysis length and BPD is biparietal diameter), EFW = 1033.286 + 12.733 x ThiM, and EFW = 1025.383 + 12.775 x ThiV. For both the formula-generating and the prospective-validation groups, there were no significant differences between the accuracies of the new 2D and 3D models in the prediction of birth weight. When applied to our population, the performances of the modified and original versions of the previously published 2D equations and the performance of the original 3D formula reported by Chang et al. were all significantly worse than our models.
We believe that the greatest sources of discrepancy in estimation of birth weight are the phenotypic differences among patients used to create each of the formulae mentioned in this study. Our data reinforce the need for customized birth-weight prediction formulae, regardless of whether 2D or 3D measurements are employed.
比较二维（2D）超声参数衍生的出生体重预测模型和三维（3D）超声成像测量的总胎儿大腿体积预测模型的准确性；并比较这些公式与之前发表的公式的性能。
共评估了 210 例患者，以创建一个公式生成组（n=150）和一个前瞻性验证组（n=60）。对第一组进行多项式回归分析，生成一个基于 2D 超声测量的方程、一个基于多平面技术（ThiM）测量的胎儿大腿体积的方程和一个基于虚拟器官计算机辅助分析（VOCAL（））方法获得的胎儿大腿体积的方程（ThiV）。使用配对样本 t 检验和 Bonferroni 调整比较这些方程在公式发现组和前瞻性验证组中的性能。采用相同的方法比较新的 2D 和 3D 公式与之前发表的原始和修改的 2D 公式以及 Chang 等人报道的 3D 模型的准确性。
预测出生体重的最佳拟合公式为：估计胎儿体重（EFW）=-562.824+11.962xACxFDL+0.009xBPD(2)xAC(2)（其中 AC 为腹围，FDL 为股骨骨干长度，BPD 为双顶径），EFW=1033.286+12.733xThiM，EFW=1025.383+12.775xThiV。对于公式生成组和前瞻性验证组，新的 2D 和 3D 模型在预测出生体重方面的准确性均无显著差异。应用于我们的人群时，之前发表的 2D 方程的修改和原始版本以及 Chang 等人报道的原始 3D 公式的性能均明显逊于我们的模型。
我们认为，估计出生体重的最大差异来源是用于创建本研究中提到的每个公式的患者的表型差异。我们的数据强化了需要制定定制的出生体重预测公式的观点，无论使用 2D 还是 3D 测量。
