Optimizing a first-trimester predictive model for Trisomy 21

Trisomy 21, commonly known as Down syndrome, is the most frequent autosomal chromosomal abnormality, occurring in 1 of every 800 live births. Recently, first-trimester screening models that combine maternal age, nuchal translucency (NT), pregnancy associated plasma protein A (PAPP-A) and free beta human chorionic gonadotropin (hCG) have reported a sensitivity for detection of Down syndrome pregnancies of 80-90% for a 5% false positive rate. An even more sensitive screening method for Trisomy 21 in the first trimester would allow improved detection and reduce the number of invasive procedures needed to confirm the diagnosis.; Methods. This exploratory study used data from the First Trimester Maternal Serum Biochemistry and Fetal Ultrasound Nuchal Translucency Screening (BUN) study to investigate whether the ultrasound and biochemical markers should be adjusted for maternal weight, race and smoking and if these adjustments would improve the overall performance of the screening model. Two risk methods for determining risk were explored, the likelihood ratio method and logistic regression. The performance of the newly developed model was then compared to the performance of the original BUN algorithm using an independent dataset.; Results. Adjusting the ultrasound and biochemical markers for weight and race increased the sensitivity to 78.7% compared with the unadjusted sensitivity of 75.4% for a 5% false positive rate when the likelihood ratio method was used to calculate risk. Smoking did not improve the overall performance of the model. Adding weight to the logistic regression model increased the detection rate of Trisomy 21 from 77.0% to 78.7%. Race and smoking did not improve the detection rate of Trisomy 21 when logistic regression was used. The new algorithm performed similarly to the original model on the BUN data. When the two screening models were compared using an independent dataset, the original BUN algorithm had a slightly better overall performance than the adjusted algorithm.; Conclusions. When used for first trimester screening, PAPP-A and free beta hCG are dependent on maternal weight and race and NT is dependent on race. The introduction of these adjustments into a risk assessment model did not significantly improve detection.