Down's syndrome screening: population statistic dependency of screening performance

Abstract

The choice of parameter sets used to calculate Down's syndrome risks is important. This study details analysis of samples from affected and unaffected pregnancies and evaluates whether published population data is optimal. Screening efficiency realized with measurement procedure-specific population parameters is compared with selected population sets available in the literature. In a retrospective experiment, double and triple testing was performed on maternal serum samples from 286 randomly chosen unaffected singleton pregnancies and 95 Down's syndrome affected pregnancy samples. Using a risk cut-off of 1 in 250, detection rates and false positive rates were estimated for different population settings to select a model giving the best overall efficacy. Receiver operation characteristics curve analysis was performed and detection rates realized with the different population settings was estimated at a 5% fixed false positive rate. Geometric mean weight corrected multiples of the median values were 1.01 for alpha-fetoprotein (AFP), 1.02 for human chorionic gonadotropin (hCG) and 1.01 for unconjugated estriol (uE3) in unaffected pregnancies and 0.77 (95% CI: 0.71-0.83) for AFP, 2.42 (95% CI: 2.17-2.71) for hCG and 0.78 (95% CI: 0.73-0.83) for uE3 in affected pregnancies. Differences in double and triple risks obtained with the different models were significantly different from each other (p < 0.001). At a cut-off of 1 in 250, the maximum triple test detection rate was 75.8% for a false positive rate of 4.9% and was obtained with the measurement procedure-specific setting. At a fixed false positive rate of 5%, the maximum detection rate for the triple test was 77.9% (95% CI: 62.2%-85.8%). The maximum double test detection rate at 5% false positive rate was 69.6% (95% CI: 59.5%-78.5%). Except for two models, the area under the curve for the triple test was higher than that of the double test. The Access triple test meets the typical performance characteristics for this test combination. The assay-specific settings yielded the overall best efficacy for the criteria studied. Therefore, the availability of measurement procedure-specific mid-trimester reference values for unaffected and affected pregnancies in prenatal screening programs is essential. Such reference values are established for the Beckman Coulter Access triple test: maternal serum AFP, uE3 and hCG.