The triple-marker test in predicting fetal aneuploidy: a compromise between sensitivity and specificity

to review the contribution of unconjugated estriol in Down's syndrome detection, and influence of maternal age, cut-off choice, and population specificity on the balance between triple-marker test sensitivity and specificity.

Prenatal karyotyping was performed in 2833 pregnant women, 73% of them over the age of 34. Duration of gestation was determined by ultrasound in 98% of women. Prior to amniocentesis, the serum levels of alpha-fetoprotein, human chorionic gonadotropin and unconjugated estriol were evaluated and corrected for weight. The risk of trisomy 21 was calculated for the first 986 subjects using default medians, and for 1847 by our population-specific medians. The cut-off was initially 1:300 at term, but the 1:100 and 1:200 risks were also tested. Down syndrome risk was calculated with alpha-fetoprotein and human chorionic gonadotropin combination as well. Linear logistic regression model was performed to test the ability of aneuploidy markers to classify patients into normal and trisomic groups.

There were twelve cases of Down's syndrome, seven of trisomy 18, four of trisomy 13, and one trisomy 22. Four cases of aneuploidy (16.7%) referred to women younger than 35. With the cut-off risk of 1:300, detection rate was 87.5% and specificity 63.3%, and with 1:100, 66.7% and 79.5%, respectively. The sensitivity for Down's syndrome was from 75% for cut-off=1:100 to 92% for cut-off=1:300, while detection of other trisomies was less successful (58% and 83%, respectively). Exclusion of unconjugated estriol MoM from the risk calculations reduced detection rate by 33% and improved specificity by 4% independently of cut-off choice. Linear logistic regression analysis showed that only unconjugated estriol was able to correctly classify patients between normal and trisomy 21 (p=0.011, odds ratio=1.445), and normal and trisomy 18 (p=0.0023, odds ratio=1.96) groups.

The unconjugated estriol significantly contributes in Down's syndrome detection with cost of slightly reduced specificity. The 1:300 risk caused an unfavorable compromise between sensitivity and specificity. A higher cut-off, 1:100, would indicate performance of amniocentesis in women aged 39 years and older, and in those aged 35-39 only after the screen-positive result.