Latent class analysis applied to patterns of fetal sonographic abnormalities: definition of phenotypes associated...

The aim of the present study was to generate different latent variables that classify the major chromosome aneuploidies using frequency and patterns of fetal sonographic abnormalities in a large database. A total of 1867 fetuses with sonographic abnormalities recorded in a database at New England Medical Center from January 1995 to March 1998 were available for the statistical analysis. Included within this group were 61 aneuploid fetuses, including 11 with 45,X, 30 with trisomy 21, 14 with trisomy 18 and 6 with trisomy 13, 40 structural malformations and/or sonographic markers were detected in these 61 aneuploid fetuses. The ability of malformations and sonographic markers to generate different groups of phenotypes was evaluated by means of latent class analysis, using the 61 affected cases. Four different classes were generated with the hypothetical assumption that each of them could satisfactorily identify a respective fetal aneuploidy represented in the study group. Among 40 fetal malformations and/or sonographic markers, the most important findings in generating specific karyotypic groups were cystic hygroma (class 1), duodenal atresia (class 2), holoprosencephaly (class 3) and omphalocele (class 4), respectively. Accuracy of the classification was 72 per cent for Turner syndrome (class 1), 74 per cent for Down syndrome (classes 1 and 2), 88 per cent for trisomy 13 (class 3) and 93 per cent for trisomy 18. The frequency of associated malformations detected sonographically can help to define a phenotype that is likely to be representative of a specific aneuploidy. Before the definitive karyotype is available or, in cases in which patients refuse an invasive prenatal diagnostic procedure, this may improve antenatal clinical management.