Normal amniotic fluid volume across gestation: Comparison of statistical approaches in 1190 normal amniotic fluid volumes. Issue 7 (15th May 2017)

• Record Type: Journal Article
• Authors: Ounpraseuth, Songthip T.; Magann, Everett F.; Spencer, Horace J.; Rabie, Nader Z.; Sandlin, Adam T.
• Is Part Of: Journal of obstetrics and gynaecology research. Volume 43:Issue 7(2017)
• Page Start: 1122
• Page End: 1131
• Publication Date: 2017-05-15
• Subjects: amniotic fluid volume assessment -- oligohydramnios/polyhydramnios -- pregnancy -- statistics; Gynecology -- Periodicals; Obstetrics -- Periodicals; 618.1005
• DOI: 10.1111/jog.13332
• Languages: English
• ISSNs: 1341-8076

Abstract

Abstract: Aim: Ultrasound estimation and evaluation of amniotic fluid volume (AFV) is an important component of pregnancy surveillance and fetal well‐being. The purpose of this study was to compare and contrast four statistical methods used to construct gestational age‐specific reference intervals for the assessment of AFV. Methods: A total of 1095 normal AFV derived from four studies that measured AFV using dye‐dilution or direct measurement at the time of hysterotomy were used to construct reference intervals using polynomial regression, quantile regression, Royston and Wright mean and SD, and Cole's lambda mu sigma (LMS) methods. The 2.5th, 5th, 50th, 95th, and 97.5th centiles were derived for each statistical method. Results: AFV increased curvilinearly from 15 gestational weeks and onward. Based on the 50th centile, the maximum value occurred at 30 weeks' gestation for the polynomial regression and mean and SD methods while the maximum was achieved at week 31 for the quantile regression and LMS methods. When data were sparse, the quantile regression method produced dramatically different estimates at the higher centile. Conclusion: The four statistical methods produced similar results at gestational ages in which AFV was high. The quantile regression approach, however, produces results that are more reflective of the data when the data are sparse. Given the flexibility and robustness of the quantile regression method, we recommend its use in constructing reference intervals when the interest lies in the tails of the reference distribution.