Evaluation of the accuracy of ultrasonography fetal weight estimation models; assessing regression formulae in a Turkish population

• Published in: Journal of obstetrics and gynaecology Vol. 35; no. 5; p. 437
• Main Authors: Özdamar, Ö, Gün, I, Öner, Ö, Özden, O
• Format: Journal Article
• Language: English
• Published: England 04.07.2015
• Subjects: Cross-Sectional Studies; Female; Fetal Weight; Humans; Infant, Newborn; Pregnancy; Regression Analysis; Retrospective Studies; Ultrasonography, Prenatal; Accuracy; actual birth weight; estimated birth weight; validity; ultrasonography
• ISSN: 1364-6893
• DOI: 10.3109/01443615.2014.968111

In this study, we aimed to evaluate the accuracy of fetal weight prediction, to investigate the validity of sonographic fetal anthropometric parameters in a Turkish population and to assess the most commonly used sonographic formulas for estimation of fetal birth weight. Our retrospective, cross-sectional study included 126 singleton deliveries between June 2010 and January 2011, at the Department of Obstetrics and Gynecology of the GATA Haydarpaşa Training Hospital. Ultrasonography measurement results were applied to nine different fetal weight estimation formulas. Mean error, mean absolute error, mean percentage error and mean absolute percentage error rates were calculated. Under- or overestimation rates and correlation coefficients were also calculated. Fetal biparietal diameter (BPD) and abdominal circumference (AC) were significantly correlated with the actual birth weight and the power analysis for both parameters was calculated as 81%. The highest correlation coefficients in our general population were those of the F1 (Hadlock 1) and F2 (Hadlock 2) models. The highest mean percentage error was detected on F8 (Merz 2) model. Fetal weight estimation modalities were observed to give the best results in a weight range of 3,000 and 3,500 g. Regression formulae used in our population, in general, tended to underestimate, however Hadlock 1 and 2 formulations provided the most accurate results. The Hadlock 1 formula estimated the closest to the actual birth weight in fetuses expected to be born under 3,000 g or over 3,500 g.