Objective
We sought to evaluate performance of a noninvasive prenatal test for fetal trisomy 21 (T21) and trisomy 18 (T18).
Study Design
A multicenter cohort study was performed whereby cell-free DNA from maternal plasma was analyzed. Chromosome-selective sequencing on chromosomes 21 and 18 was performed with reporting of an aneuploidy risk (High Risk or Low Risk) for each subject.
Results
Of the 81 T21 cases, all were classified as High Risk for T21 and there was 1 false-positive result among the 2888 normal cases, for a sensitivity of 100% (95% confidence interval [CI], 95.5–100%) and a false-positive rate of 0.03% (95% CI, 0.002–0.20%). Of the 38 T18 cases, 37 were classified as High Risk and there were 2 false-positive results among the 2888 normal cases, for a sensitivity of 97.4% (95% CI, 86.5–99.9%) and a false-positive rate of 0.07% (95% CI, 0.02–0.25%).
Conclusion
Chromosome-selective sequencing of cell-free DNA and application of an individualized risk algorithm is effective in the detection of fetal T21 and T18.
Currently, the most effective and commonly used prenatal screening tests for fetal aneuploidy use a combination of maternal age, sonographic measurement of the fetal nuchal translucency, and measurement of maternal serum screening markers in the first and second trimesters. Although prenatal screening tests have greatly improved in the past decade, the best performing screening tests have false-positive rates of 2-3% and false-negative rates of ≥5%. Positive screening results require confirmation with diagnostic testing (eg, chorionic villus sampling [CVS] or amniocentesis); these tests carry fetal loss rates of approximately 1 in 300 procedures. Current screening paradigms are not uniform, with multiple algorithms available for use at various stages of pregnancy, and therefore can be confusing to incorporate into clinical practice.
For Editors’ Commentary, see Contents
The presence of fetal and maternal cell-free DNA (cfDNA) circulating in maternal plasma is now widely appreciated, and several groups have demonstrated fetal trisomy 21 (T21) detection using massively parallel DNA shotgun sequencing (MPSS) in case-control studies. This technique sequences cfDNA fragments to determine their specific chromosomal origin; a slightly higher than expected percentage of chromosome 21 fragments indicates that the fetus has a third chromosome 21. In addition to detecting T21, several studies have reported on the use of MPSS in assaying trisomy 18 (T18) and trisomy 13 (T13).
Despite these promising results, MPSS randomly analyzes DNA from the entire genome, resulting in higher cost and complexity than is practical for widespread clinical adoption. Recent studies have reported on an alternative assay, Digital ANalysis of Selected Regions (DANSR), that selectively evaluates specific genomic fragments from cfDNA, providing more efficient use of sequencing and potentially reduced costs when compared to MPSS. This process of chromosome-selective sequencing has been extended to enable simultaneous determination of the fraction of fetal cfDNA in the maternal plasma as well as the chromosome proportion by assaying polymorphic and nonpolymorphic loci. When combined with a novel analysis algorithm, the Fetal-fraction Optimized Risk of Trisomy Evaluation (FORTE), this information can provide an individualized assessment of trisomy risk. In a recently published blinded independent study, the use of DANSR and FORTE was found to separate all cases of T21 and 98% of cases of T18 from euploid pregnancies in 400 singleton pregnancies at 11-13 weeks’ gestation.
This report describes the results of a multicenter study designed to evaluate the performance of this noninvasive prenatal assay and algorithm in a large cohort of women prior to invasive prenatal diagnostic testing.
Materials and Methods
Study population
This was a prospective, cohort study comprising pregnant women aged ≥18 years, at gestational age ≥10 weeks, with a singleton pregnancy, who were planning to undergo invasive prenatal diagnosis for any indication. Subjects who were pregnant with >1 fetus, or who themselves had a known aneuploidy, had active malignancy or a history of metastatic cancer, or had already undergone CVS or amniocentesis during the current pregnancy were excluded.
Subjects were prospectively enrolled after providing informed consent at selected prenatal care centers in the United States, The Netherlands, and Sweden. Institutional review board approval was obtained at all participating centers.
Sample collection and preparation
Approximately 20 mL of blood was collected from each subject prior to any invasive procedure into a Cell-free BCT tube (Streck, Omaha, NE). Samples were sent directly to the laboratory without processing and needed to be received within 7 days of collection with no temperature excursions indicating freezing. Plasma was isolated from blood via a double centrifugation protocol. cfDNA was isolated from plasma using the Dynabeads Viral NA DNA purification kit (Dynal, Grand Island, NY) protocol, with minor modifications, and each sample was arrayed into individual wells of a 96-well microtiter plate.
Test methods
Each subject’s cfDNA sample was isolated and quantified using the DANSR assay, which has been described previously. Briefly, this method uses ligation of locus-specific oligonucleotides to produce a sequencing template only from selected genomic loci, thus reducing the amount of DNA sequencing needed. The FORTE algorithm, also previously described in detail, was used to estimate the risk of aneuploidy for chromosomes 21 and 18 in each sample. The FORTE risk score is determined by calculating the odds ratio for trisomy based on chromosome 21 and 18 cfDNA counts, and fraction of fetal cfDNA in the sample, then applying this as a likelihood ratio to the a priori trisomy risk based on the maternal age and gestational age. A predefined cutoff value of 1 in 100 (1%) was designated as the threshold for classifying a sample as High Risk vs Low Risk. The cutoff value was determined based on previous analyses that demonstrated an optimal separation between trisomy and euploid samples. Samples that did not generate a result were classified as low (<4%) fraction of fetal cfDNA, inability to measure fraction of fetal cfDNA, unusually high variation in cfDNA counts, and failed sequencing.
The laboratory personnel who performed the analyses were blinded to the clinical information associated with each sample. Finalized results were transferred to an independent data management center (Advance Research Associates, Mountain View, CA) for merging of assay and clinical data, and unblinding.
Data analysis
Sample size was calculated based on obtaining sufficient cases of T21 to achieve lower bound 95% confidence intervals (CI) for sensitivity and specificity that were comparable or superior to current prenatal screening tests. The target performance for the DANSR and FORTE method was anticipated to be ≥98% for both sensitivity and specificity based on previous data. Using this estimate, at least 60 cases of T21 would be required to provide a lower 95% CI of 90% for sensitivity. Assuming a T21 prevalence of 1 in 50 in the study cohort, based on a typical population of women undergoing invasive prenatal diagnosis, at least 3000 eligible subjects would be required. Categorical variables were summarized by the number and percentage of subjects in each category. Continuous variables were summarized as total number, mean, SD, minimum, median, and maximum values. We used χ 2 tests with Bonferroni correction when comparing categorical variables and proportions. Linear regression models were used to test the correlation between continuous variables (eg, percent fetal and gestational age) with the null hypothesis that the slope between 2 continuous variables is 0. Multivariate logistic regression was used when the response variable was categorical. Standard analysis of variance models were used when the response variable was continuous and the explanatory variables were discrete. The 95% CI for sensitivity and specificity were computed using the method of Wilson.
Analysis of samples using DANSR and FORTE included all evaluable subjects who had undergone invasive testing with fetal genotype analysis by karyotype, fluorescent in situ hybridization, or quantitative fluorescent polymerase chain reaction.
Prior to study unblinding, chromosomal abnormalities were categorized as T21, T18, T13, sex chromosome aneuploidy, triploidy, balanced translocation, unbalanced translocation, duplication, deletion, extra structurally abnormal chromosome, confined placental mosaic, mosaic (likely true), mosaic (likely artifact), and other. T21 cases resulting from unbalanced Robertsonian translocations were classified as “T21” and those with mosaicism for T21 or T18 were classified as “other.” Subjects with commonly identified chromosomal rearrangements predicted to have a normal outcome, such as inversions of chromosomes 1, 9, or 16, in addition to those with balanced Robertsonian translocations (inherited and de novo) were considered normal for the purpose of this analysis. Categorization was performed by a clinical geneticist (M.E.N.) based on review of each clinical karyotype report for all abnormal results.
For efficacy analysis, results from the DANSR assay and FORTE algorithm were compared against the reference standards of clinically adjudicated invasive testing results. Results from the DANSR assay and FORTE algorithm were provided as a trisomy risk score, with the upper and lower risk value capped at 99% (99 in 100) and 0.01% (1 in 10,000), respectively. Primary calculations for sensitivity and specificity were based on a 1% (1 in 100) cutoff to designate results as High Risk or Low Risk for a given trisomy. Additional analyses on test performance were reported at different risk cutoff values.
Results
Study participants
A total of 4002 pregnant women were enrolled in the study from Aug. 1, 2010, through Nov. 1, 2011, across 3 countries. Samples from all 3 countries were analyzed together as a single cohort given the use of specialized blood collection tubes that preserve cfDNA in blood for up to 14 days. Of the 4002 plasma samples obtained, 433 samples were used for assay development with a subset of these reported on previously. An additional 341 samples were ineligible prior to analysis for failing to meet inclusion/exclusion criteria (n = 237), insufficient sample volume (n = 84), and incorrect sample labeling (n = 20). Of the 3228 samples remaining, all underwent analysis. Of analyzed samples, 57/3228 cases (1.8%) were excluded due to low (<4%) fraction of fetal cfDNA and an additional 91/3228 (2.8%) samples were excluded due to assay failure. Assay failure modes included inability to measure fraction of fetal cfDNA, unusually high variation in cfDNA counts, and failed sequencing. Figure 1 shows the flow of samples between enrollment and analysis.
The mean maternal age was 34.3 (range, 18–50) years for all subjects, and the cohort was racially and ethnically diverse, with 49.6% (1600/3228) of the population identified as Caucasian, 6.4% (207/3228) African American, 13.4% (434/3228) Asian, 22.7% (732/3228) Hispanic, and 7.9% (255/3228) other. For invasive testing, 818/3228 (25.3%) of subjects had karyotyping with CVS, and 2410/3228 (74.7%) by amniocentesis. There were no statistical differences as determined by χ 2 tests with Bonferroni correction between the normal and trisomy groups among these variables.
The mean gestational age of the cohort was 16.9 (range, 10.0–38.7) weeks with no statistical difference between normal and trisomy groups based on linear regression analysis ( Table 1 ). In the entire cohort for analysis, there were 84 cases of T21, 42 cases of T18, and 81 other abnormal karyotypes including 4 cases of T13 and T21 sex chromosomal aneuploidies ( Table 2 ).
| Demographic | Normal a (n = 3021) | Trisomy 21 (n = 84) | Trisomy 18 (n = 42) | Other (n = 81) | Total (n = 3228) |
|---|---|---|---|---|---|
| Maternal age, y, mean ± SD (range) | 34.3 ± 6.3 (18–50) | 35.4 ± 7.3 (18–47) | 34.5 ± 6.1 (22–45) | 32.0 ± 6.9 b (18–44) | 34.3 ± 6.4 (18–50) |
| Gestational age, wk, mean ± SD (range) | 17 ± 4.1 (10–38.7) | 16.4 ± 3.1 (11.6–25.7) | 16.2 ± 4.3 (10.9–29.4) | 16.6 ± 4.5 (10.0–34.1) | 16.9 ± 4.1 (10–38.7) |
| Maternal ethnicity, n (%) | 1504 (49.8) | 36 (42.9) | 20 (47.6) | 40 (49.4) | 1600 (49.6) |
| Caucasian | 197 (6.5) | 3 (3.6) | 2 (4.8) | 5 (6.2) | 207 (6.4) |
| African American | 406 (13.4) | 8 (9.5) | 5 (11.9) | 15 (18.5) | 434 (13.4) |
| Asian | 677 (22.4) | 27 (32.1) | 10 (23.8) | 18 (22.2) | 732 (22.7) |
| Hispanic | 237 (7.8) | 10 (11.9) | 5 (11.9) | 3 (3.7) | 255 (7.9) |
| Other | |||||
| Fetal DNA in sample | |||||
| n c | 2888 | 81 | 38 | 73 | 3080 |
| Percent fetal DNA, mean ± SD (range) | 11 ± 4.5 (4.2–51.3) | 11.6 ± 4.2 (5.1–23.3) | 10 ± 3.8 (4.9–20.8) | 11.6 ± 5.1 (4.5–32.0) | 11 ± 4.5 (4.2–51.3) |
a For purpose of this analysis, “normal” includes common chromosomal variants and balanced Robertsonian translocations;
| Chromosomal abnormality type | n | Percent of all abnormalities |
|---|---|---|
| Trisomy 21 | 84 | 40.6 |
| Trisomy 18 | 42 | 20.3 |
| Trisomy 13 | 4 | 1.9 |
| Sex chromosome aneuploidy a | 21 | 10.1 |
| Triploidy | 5 | 2.4 |
| Balanced translocation | 10 | 4.8 |
| Unbalanced translocation | 3 | 1.4 |
| Duplication | 3 | 1.4 |
| Deletion | 2 | 1.0 |
| Extra structurally abnormal chromosome | 5 | 2.4 |
| Confined placental mosaic | 8 | 3.9 |
| Mosaic–likely true | 11 | 5.3 |
| Mosaic–likely artifact | 3 | 1.4 |
| Other | 6 | 2.9 |
Fetal fraction
Overall, the fraction of fetal cfDNA expressed as a percentage was 11% in the samples tested ( Table 1 ). The fetal fraction did not vary with race/ethnicity, maternal age, or trisomy type using standard 1-way analysis of variance and linear regression analysis. The fraction of fetal cfDNA by gestational age week is depicted in Figure 2 . For gestational ages between 10-22 weeks, there was no statistical difference in fraction of fetal cfDNA.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
