Can reliable Down's syndrome detection rates be determined from prenatal screening intervention trials?

OBJECTIVES - To develop a standardised approach for analysing Down's syndrome screening performance in clinical practice and to apply it to published intervention trials in order to estimate detection and false positive rates more accurately. METHODS - Peer reviewed intervention trials, grouped by specific combination of analytes, were reanalysed. Revised detection rates were calculated for each study, taking into account both the high spontaneous loss during the last half of pregnancy and the possible under ascertainment of Down's syndrome live births not detected by screening. Collective screening performance was estimated, when possible, using a published methodology based on fitting receiver-operator characteristic curves. RESULTS - Sixteen trials were analysed; 11 using three, and five using two, analytes. Collective screening performance for the triple analyte trials was Down's syndrome detection rates of 57, 64, and 69% at amniocentesis referral rates of 3, 5, and 7% respectively. Four of the five studies involving two analytes performed less well, individually, when compared with the overall performance of the three analyte studies. It was not possible to estimate collective performance for the two analyte studies because there were too few. CONCLUSIONS - Accurate Down's syndrome detection rates are difficult to obtain in intervention trials owing to two potential biases, both of which tend to produce overestimates of the true rates. These sources of bias need to be taken into account when analysing and reporting Down's syndrome intervention trials. The methodology presented here offers the opportunity to achieve a more reliable, standardised estimate of both individual and collective intervention trial screening performance.