Background: Normative ranges of fetal echocardiographic measurements are important for quantitative diagnosis of fetal cardiovascular disease. The current normative ranges were derived from small samples and were based on the hypothesis of a normal distribution of these measurements during fetal cardiovascular growth. The aims of this study were to test the hypothesis of a normal distribution of fetal echocardiographic measurements in a large multicenter cohort and to propose a reference system without the normal distribution hypothesis to improve accuracy of fetal echocardiographic measurements.
Methods: Fifty-two variables from 6,343 normal fetal echocardiographic examinations were acquired from seven Chinese centers. The hypothesis of a normal distribution used in ordinary least squares regression was tested with the Jarque-Bera test. The quantile score (q score) derived from quantile regression without normal distribution hypothesis was compared with the Z score derived from ordinary least squares regression. A total of 288 fetuses with outflow tract and great artery abnormalities and 300 normal fetuses were used to compare the diagnostic accuracy of q and Z scores.
Results: All fetal echocardiographic measurements showed non-normal distributions (P < .001). The normal range was underestimated by ordinary least squares regression compared with quantile regression by 30 ± 11%. The partial normalized areas under the receiver operating characteristic curve within the 20% false-positive rate were 0.62 and 0.50 for the q and Z scores, respectively.
Conclusions: The q score provides a more robust system for determining normative ranges of fetal echocardiographic measurements. The improved sensitivity of matched false-positive rates makes the q score a more accurate reference for prenatal diagnosis, assessment, and prognosis of fetal cardiovascular disease.
Keywords: Fetal echocardiography; Quantile score; Z score.
Copyright © 2018 American Society of Echocardiography. All rights reserved.