Premium
Comparison of likelihood ratios of first‐trimester nuchal translucency measurements: multiples of median, delta or mixture
Author(s) -
Sahota D. S.,
Leung T. Y.,
Chen M.,
Chan L. W.,
Fung T. Y.,
Lau T. K.
Publication year - 2010
Publication title -
ultrasound in obstetrics and gynecology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 3.202
H-Index - 141
eISSN - 1469-0705
pISSN - 0960-7692
DOI - 10.1002/uog.7677
Subject(s) - trisomy , medicine , intraclass correlation , nuchal translucency , obstetrics , gynecology , statistics , first trimester , pregnancy , nuclear medicine , mathematics , reproducibility , fetus , biology , genetics
Objective To compare the likelihood ratios (LR) determined for a measured nuchal translucency (NT) using the multiples of the median (MoM), delta‐NT and mixture model approaches. Methods This was a retrospective study of singleton pregnancies with known outcome screened at the 12 th week of pregnancy between January 2006 and May 2008. NT, crown–rump length (CRL) and LR for trisomy 21 using the delta‐NT approach (LRNT‐delta) were extracted from The Fetal Medicine Foundation trisomy 21 risk‐calculation software database. NT measurements were converted into their equivalent MoM by dividing by the expected CRL‐specific median NT. LR for trisomy 21 based on the MoM (LRNT‐MoM) of each case was then calculated as the ratio of the Gaussian probability density functions using published trisomy 21 risk‐screening parameters. The LR using the mixture model (LRNT‐mixture) was calculated as the ratio of individual composite probability densities reported for trisomy 21 and unaffected pregnancies. The intraclass correlation coefficient (ICC) was used to assess the agreement between the three different approaches while the paired t ‐test was used to perform pairwise comparisons. Results Of the 5907 pregnancies with known outcome, 14 had trisomy 21. The ICC between the three approaches to determining an LR was 0.947 (95% CI, 0.943–0.951; P < 0.0001). Pairwise comparison indicated that differences between LR were statistically significant ( P < 0.0001). Likelihood ratios determined via the delta‐NT model were in general greater than those produced by both the mixture model ( z = 5.58, P < 0.0001) and the MoM approaches ( z = 32.8, P < 0.0001). For a 5% false‐positive rate the detection rate of trisomy 21 (DR) was 78.6% (95% CI, 57.1–99.9%) The screen positive thresholds to achieve this DR were 1 : 200 for the delta‐NT approach, 1 : 220 for the mixture approach and 1 : 290 for the MoM approach. Conclusion No single method of determining NT‐associated LR outperformed the others in its ability to detect a trisomy 21‐affected pregnancy. Copyright © 2010 ISUOG. Published by John Wiley & Sons, Ltd.