Aims: Risk estimation for Down's syndrome in antenatal serum screening with maternal age and multiple serum biomarkers is usually complicated and computationally intensive. We have developed a simple scoring system using the Spiegelhalter-Knill-Jones approach, which was based on Bayesian theorem and the logistic regression model.
Methods: A prospective data set with 3842 singleton pregnancies including 6 affected pregnancies served as "trained data". Maternal age, maternal serum alpha-fetoprotein and human chorionic gonadotrophin levels of each pregnant woman were adopted as the predictors to establish the scoring model using the S-KJ approach. Model validation was undertaken using a receiver operating characteristics (ROC) curve with another 3050 singleton pregnancies including 4 affected pregnancies ("validated data").
Results: For the trained data the sensitivity and specificity of the scoring system at cut-off value of 1:250 was 66.7% and 92.6%, respectively. For the validated data the sensitivity and specificity at the same cut-off point was 75% and 92.2%, respectively. The area under the ROC curve of the trained and validated data was 76.96% (95% CI: 51.80-100%), and 94.07% (95% CI: 84.47-100%), respectively.
Conclusions: The S-KJ scoring system has been demonstrated to be a simple, and efficient method for the risk estimation of Down's syndrome. This system can be applied to other antenatal serum screening systems.