Mid trimester sonographic findings for the prediction of Down syndrome in a sonographically screened population

Objective The purpose of this study was to report the adjusted effect sizes of mid trimester sonographic findings that are associated with Down syndrome in a sonographically screened population. Study design A large prospective single-center cohort study was conducted between March 1993 and December 2002 in South-East Queensland with women who were first scanned between 15 to 22 weeks of gestation. Univariate and multivariable logistic regression modeling was used to relate karyotypically ascertained Down syndrome fetuses and their control counterparts against routinely collected demographic and sonographic findings. Results Data were available for 73 Down-affected and 16,891 unaffected pregnancies. Strong colinearity existed between short humerus and short femurs that necessitated the removal of FL in pursuant multivariable models. In the most parsimonious model, which was adjusted for maternal age and gestational age, pregnancies with thick nuchal skinfold (regression coefficient β [± SE], 2.100 ± 0.545), short humerus length (regression coefficient β, 2.304 ± 0.314), presence of echogenic bowel (regression coefficient β, 1.602 ± 0.412), presence of echogenic intracardiac focus (regression coefficient β, 1.975 ± 0.308), presence of renal pelvic dilation (regression coefficient β, 1.281 ± 0.420), presence of aneuploid associated anomalies (regression coefficient β, 4.473 ± 0.535), the interaction between gestational age and thick nuchal skinfold (regression coefficient β, 0.465 ± 0.210), and the interaction between short humerus length and the presence of aneuploid associated anomalies (regression coefficient β, −1.693 ± 0.811) all were associated significantly with Down syndrome risk (all P< .05). Adjusted relative risk estimates were substantially different from their crude estimates. Conclusion Routinely collected mid trimester sonographic findings are associated significantly with Down syndrome risk in a sonographically screened population after accounting for maternal age and gestational age. Because of dependencies between ultrasonic findings, risk estimates should be derived from appropriate multivariable models.