Down's syndrome epidemiology and risk estimation

Abstract

Prenatal screening for Down's syndrome uses a risk estimate as the screening variable. Modifying the age-associated risk by measuring biochemical markers or fetal measurements requires more sophisticated mathematical techniques than calculating the observed estimates of the likelihood ratio. Frequently, the screening performance is reported on the same data used to generate the model. This can lead to an overestimate of the performance likely to be achievable in practice. In order to determine whether a proposed model for screening is better (or worse) than an established model, the two should be directly compared on a reasonably sized dataset of cases and controls that was not used to derive either of the models. At best, only marginal gains in screening performance can be expected using new or refined models for assigning Down's syndrome risk.