Accuracy of the Ultrasound Estimate of the Amniotic Fluid Volume (Amniotic Fluid Index and Single Deepest Pocket) to Identify Actual Low, Normal, and High Amniotic Fluid Volumes as Determined by Quantile Regression

Objectives To identify abnormal amniotic fluid volumes (AFVs), normal volumes must be determined. Multiple statistical methods are used to define normal amniotic fluid curves; however, quantile regression (QR) is gaining favor. We reanalyzed ultrasound estimates in identifying oligohydramnios, normal fluid, and polyhydramnios using normal volumes calculated by QR. Methods Data from 506 dye-determined or directly measured AFVs along with ultrasound estimates were analyzed. Each was classified as low, normal, or high for both the single deepest pocket (SDP) and amniotic fluid index (AFI). A weighted kappa statistic was used to assess the level of agreement between the AFI and SDP compared to actual AFVs by QR. Results The overall level of agreement for the AFI was fair (kappa = 0.26), and that for the SDP was slight (kappa = 0.19). Although not statistically significant (P = .792), the positive predictive value to classify a low volume using the AFI was lower compared to the SDP (35% vs 43%). The positive predictive value for a high volume was higher using the AFI compared to the SDP (55% versus 31%) but not statistically significant. The missed-call rate for high-volume identification by the SDP versus AFI was statistically significant (odds ratio, 5.5; 95% confidence interval, 2.04-14.97). The missed-call rate for low-volume identification by the AFI versus SDP was not statistically significant (odds ratio, 3.3; 95% confidence interval, 0.96-11.53). Conclusions Both the AFI and SDP identify actual normal AFVs by QR, with sensitivity higher than 90%. The SDP is superior for identification of oligohydramnios, and the AFI superior for identification of polyhydramnios.