Down's syndrome risk estimates demonstrate considerable heterogeneity despite homogeneity of input.

Abstract

BACKGROUND Due to concerns about screening quality, Down's syndrome screening laboratories were surveyed to identify the range of variation in risk-calculation software. METHODS All UK Down's syndrome screening laboratories were sent the confidential survey via the National External Quality Assessment Scheme. This covered the software used, its origin, the risk calculation methodology and the Gaussian population parameters. Laboratories were also given multiples of the median (MoM) data from five representative cases and asked to calculate Down's syndrome risks on their systems. RESULTS Most parameter sets could be traced to published literature. The range of risk results for identical patient data was wide; the largest differences between lowest and highest risk for a test MoM set were: alpha-fetoprotein (AFP)+human chorionic gonadotrophin (hCG), 1:95 to 1:388 = 408%; AFP+hCG+urine estriol (UE(3)), 1:2011 to 1:7000 = 348%; AFP+free-beta-hCG, 1:280 to 1:1681 = 600%; AFP+free-beta-hCG+UE(3), 1:340 to 1:13000 = 3823%. Two explanations for this variation - the prior age risk result (half-year correction) and the population parameters - are described. CONCLUSIONS (a) All laboratories should apply half-year correction for maternal age risk calculation. (b) Triple-test parameter variations result in huge variation of risk estimates, and a single national risk threshold of 1 in 250 will be unlikely to improve screening equity unless attention is also paid to standardizing population parameters. Down's syndrome screening has been controversial since it was introduced, and the possibility that more central control (including listing acceptable population parameters) may be necessary to ensure comparability of results is certain to ensure that this controversy continues.

Cite this paper

@article{Reynolds2004DownsSR, title={Down's syndrome risk estimates demonstrate considerable heterogeneity despite homogeneity of input.}, author={Timothy M Reynolds and Andy R Ellis and Rick Jones}, journal={Annals of clinical biochemistry}, year={2004}, volume={41 Pt 6}, pages={464-8} }