Diagnostic accuracy of non-invasive prenatal sex determination: a large-scale study.

Non-invasive prenatal determination of fetal sex offers a promising alternative to invasive diagnosis of X-linked diseases and fetal disorders of ambiguous genitalia (1–4). Using free-fetal DNA (ffDNA) from maternal plasma (5) and different detection systems for Y chromosome (6–9), researchers achieved high diagnostic accuracy in fetal sex prediction during the first trimester of pregnancy without excluding variable percentage of false positives (FP) and false negatives (FN) (10, 11). We previously showed higher performance of DYS14 than SRY detection system without reaching an optimal diagnostic accuracy because of the low percentage of FN and FP (12). The aim of this study was to enhance the performance of our DYS14 detection system by introducing new key elements, increasing the volume of maternal plasma for DNA extraction and employing innovative interpretation criteria of results. Finally, we audited the overall diagnostic accuracy on a large-scale study to verify whether the new protocol ensured the correct fetal sex determination. We analyzed, in two different phases, plasma samples from 513 women at 10–15 weeks of gestation. During the first non-blinded phase, we established the best threshold value (TV) to discriminate male and female fetuses and, in the second blinded phase, we applied it to assess its diagnostic performance. Fetal sex was verified with the analysis of karyotype or confirmed with phenotype at birth. The study was approved by our Regional Ethical Committee. Genomic DNA was extracted from maternal plasma (500 μl in the first and 1000 μl in the second phase) using QIAmp DSP Virus kit (Qiagen, Hilden, Germany) and analyzed by real-time polymerase chain reaction (PCR) 7300 detection system (Applied Biosystems, Foster City, CA) using DYS14 detection system to measure the quantity of ffDNA (three replicates for each maternal sample) and telomerase reverse transcriptase detection system as a quality control. Real-time PCR [quantitative PCR (qPCR)] reaction was set up as previously described (12). Results were expressed as median values with range for a descriptive statistics and analyzed by using receiver operating characteristic (ROC) curves, calculated by spss software 17.0 (SPSS Inc., Chicago, IL), to set the TV, in terms of ffDNA concentration and number of DYS14positive replicates. Karyotype or phenotype at birth revealed that among the 115 pregnant women analyzed in the first phase of study, 55 delivered one daughter and 60 one son. By evaluation of qPCR results (Table 1), we set our first TV by analysis of ROC curve calculated using the number of DYS14-positive replicates (data not shown). However, the TV was not satisfactory, except for samples with 0, 1 or 2 DYS14-positive replicates that were clearly identified as female fetus. Only one sample from a male pregnancy had 0 DYS14-positive replicates. The critical point was the interpretation of those results with three DYS14-positive replicates, as they occurred in either female or male pregnancies. Two samples from female pregnancy gave three positive replicates but with a very low ffDNA concentration, whereas all the samples from male pregnancy gave three DYS14-positive replicates with elevated ffDNA concentration (Table 1). Therefore, we built a new ROC curve combining ffDNA concentration and the number of DYS14positive replicates and selecting only those samples with three DYS14-positive replicates (2 from female and 59 from male pregnancies). We individuated the best ffDNA concentration value as 1.42 GE/ml as it allowed us to reach a 100.00% diagnostic sensitivity [95% (confidence interval) CI: 93.9–100.0] and 100.00% diagnostic specificity (95% CI: 19.3–100.0) (Fig. 1). In the second phase of study, karyotype or phenotype at birth revealed that 208 delivered a son and 190 a daughter.