Evaluation of screening methods for Down's syndrome using bootstrap comparison of ROC curves

This paper concerns the prediction of fetal Down's syndrome in pregnant women. Down's syndrome is the most common congenital cause of severe mental retardation. We elaborate two predictive functions of trisomy 21, combining maternal age and a maternal serum marker. We evaluated them by means of receiver operating characteristic (ROC) curves which give a representation of sensitivity and specificity of a prediction model when varying the cutoff of the predictor on the whole spectrum. Since normal statistical methods for comparison of ROC curves rely on distributional assumptions which were not verified, we used bootstrapping of ROC curves as a check for the statistical significance of differences between the areas under the curves.