Sequencing-Based Tests to Determine Fetal Down Syndrome (Trisomy 21) from Maternal Plasma DNA
Fetal chromosomal abnormalities occur in approximately 1 in 160 live births. The majority of fetal chromosomal abnormalities are aneuploidies, defined as an abnormal number of chromosomes. The trisomy syndromes are aneuploidies involving 3 copies of one chromosome. Trisomy 21 (Down syndrome) is the most common form of fetal aneuploidy that is associated with survival to birth and beyond. Trisomy 18 (Edwards syndrome) and trisomy 13 (Patau syndrome) are the next most common fetal aneuploidy syndromes associated with survival to birth, although the percent of cases surviving to birth is low and survival beyond birth is limited. The most important risk factor for trisomy 21, 18, or 13 is maternal age, with an approximate risk of 1/1,600 at age 15 that increases to 1/28 by age 45.
Current guidelines recommend that all pregnant women be offered noninvasive screening for trisomy 21 before 20 weeks of gestation, regardless of age. Contemporary screening programs may also detect trisomy 18 or 13. Combinations of maternal serum markers and fetal ultrasound done at various stages of pregnancy are used, but there is not one standardized approach. The detection rate for various combinations of noninvasive tests ranges from 60–96% when the false-positive rate is set at 5%. Noninvasive screening tests are not sufficiently accurate to diagnose a trisomy syndrome and confirmatory testing is required. In addition, because of the imperfect parameters of noninvasive screening strategies, some cases will be missed and the majority of patients who are recommended to have a confirmatory invasive procedure do not have a fetus with a trisomy syndrome.
Direct karyotyping of fetal tissue obtained by invasive amniocentesis (second trimester) or chorionic villous sampling (CVS; first trimester) is required to confirm the diagnosis of trisomy. Both amniocentesis and CVS are invasive procedures and have a small but finite risk of miscarriage. A new screening strategy that reduces unnecessary amniocentesis and CVS procedures (and thus associated miscarriage) and increases detection of trisomy 21 in particular, and potentially trisomy 18 and 13 as well, has the potential to improve outcomes.
Cell-free DNA fragments can be detected in plasma of pregnant women. As early as 8 to 10 weeks of gestation, fetal DNA fragments (actually derived from the cytotrophoblastic cell layer of the placenta) comprise 6 to 10% or more of the total cell-free DNA in a maternal plasma sample. Massively parallel sequencing (MPS; also known as next-generation or “next-gen” sequencing) can be used to design assays for prenatal detection of trisomy 21; the first proof of principle studies were published in 2008. DNA fragments are first amplified by polymerase chain reaction (PCR); during the sequencing process, the amplified fragments are spatially segregated and sequenced simultaneously in a massively parallel fashion. Sequenced fragments can be mapped to the reference human genome to obtain numbers of fragment counts per chromosome. Alternatively, chromosome-targeted sequencing can be used, which obviates the need for mapping to the reference human genome.
The sequencing-derived percent of fragments from the chromosome of interest reflects the chromosomal representation of the maternal and fetal DNA fragments in the original maternal plasma sample. Additionally, in a euploid individual with a normal number of chromosomes (e.g. the woman from whom the plasma sample was taken), the proportional contribution of DNA sequences per chromosome correlates with the relative size of each chromosome in the human genome. Any detectable difference from the euploid mean for each chromosome of interest is determined for the sample. A predetermined cutoff identifies trisomy 21 or any other abnormal chromosome number. Thus, the technology must be sensitive enough to detect a slight shift in DNA fragment counts among the small fetal fragment representation of an aneuploid chromosome against a large euploid maternal background.
The overall objective of this Assessment is to determine whether nucleic acid sequencing-based testing for trisomy 21 using maternal serum improves outcomes of pregnancies screened for trisomy 21, compared to traditional serum and ultrasound testing strategies. Additionally, the evidence supporting similar objectives for trisomy 18 and 13 will be reviewed.
MEDLINE® (via PubMed) and EMBASE medical literature databases were searched for articles published in the last 5 years; limited to English-language publication is in human populations. The search was updated February 26, 2013. Several search terms were combined, such as “trisomy,” “aneuploidy,” “sequencing,” “prenatal diagnosis,” “chromosome 21” [or 18 or 13], “cell-free DNA,” etc.
Included studies had the following characteristics: 1) performed maternal plasma fetal DNA testing of pregnant women being screened for trisomy 21, trisomy 18, and trisomy 13; 2) used the final, ‘locked-down’ version of the sequencing assay that is clinically available and applied all clinical laboratory quality control measures; 3) compared the results of plasma fetal DNA testing with the results of karyotype analysis (or fluorescence in situ hybridization [FISH] if karyotype is not possible in individual cases), or with phenotype at birth; and, 4) reported information on sensitivity and specificity, or provided sufficient information to calculate these parameters.
The sensitivity and specificity estimates of sequencing-based testing for trisomy 21 were uniformly high, ranging from 99.1% to 100%, and from 99.7% to 100%, respectively. Negative predictive values, whether calculated for average (pregnant women electing screening) or high-risk (age >35) populations, were uniformly high, near or at 100% as is desirable for a screening test. Positive predictive values were 83% and 55% for high- and average-risk populations, respectively, using point estimates for test sensitivity and specificity. For trisomy 18, the sensitivity ranged from 97.2% to 100% and the specificity ranged from 99.7% to 100%. For trisomy 13 three studies reported sensitivities of 78.6–91.7%, and specificities were 99.1–100% based on a small number of cases.
A simple decision model was constructed to compare the health outcomes of nucleic acid sequencing-based testing with standard testing for trisomy 21. The strategies tested in the model include:
1. A traditional screening test followed, if positive, with an invasive procedure (CVS in the first trimester or amniocentesis in the second trimester) for confirmatory karyotyping; traditional screening tests chosen for comparison are:
a. Combined screen (first trimester, includes nuchal translucency ultrasound)
b. Integrated screen (first + second trimester serum testing and nuchal translucency ultrasound)
2. Nucleic acid sequencing-based testing in place of traditional serum screening; if positive, confirm with invasive procedure and karyotyping.
3. A traditional screening test (first trimester combined screen or first and second trimester integrated screen); positives followed with sequencing-based testing; subsequent positives confirmed by invasive procedure and karyotyping.
4. A traditional screening test with performance parameters chosen to allow better case detection along with an increased false-positive rate; positives followed with sequencing-based testing; subsequent positives confirmed by invasive procedure and karyotyping.
The outcomes of interest for this decision tree are the number of cases of trisomy 21 correctly identified, the number of cases missed, the number of invasive procedures potentially avoided (because of normal DNA test results) and the number of miscarriages potentially avoided as a result. The results were calculated for a high-risk population of women age 35 or older, and for an average-risk population including women of all ages electing an initial screen. For women testing positive on initial screen and offered an invasive, confirmatory procedure, it was assumed that only a proportion would accept, according to risk level. Sensitivities and specificities for both standard and sequencing-based screening tests were varied to represent the range of possible values.
For either high- or average-risk populations, the second strategy, screening by sequencing-based assay followed by confirmatory testing, detects the most cases. Base case estimates show detection of nearly the maximum possible number of cases for both populations. The improvement over first or second trimester standard screening assays is approximately 3–16% (base case). At the same time, the number of invasive procedures needed is reduced by as much as 80%. The number of total miscarriages after an invasive confirmatory procedure in an average risk population is also reduced from, for example, the 22 per 100,000 seen after integrated screening to 4 (an 82% reduction) using base case estimates. Confidence in negative results is high as no more than 10 of 100,000 (0.01%) screens are false negatives using base case estimates.
When added after a positive traditional screen, a sequencing-based assay does not improve the trisomy 21 case detection rate. However, fewer invasive procedures are needed than after screening by sequencing alone, due partially to the lower case detection rate. The number of total miscarriages is reduced to similar or slightly lower numbers than screening by sequencing alone. These results are seen for both high- and average-risk populations.
Re-interpretation of a first trimester traditional combined test using altered parameters allows an increased detection rate. Following positive results from such an increased sensitivity combined assay with a sequencing-based assay could take advantage of the increased detection rate, reduce the number of sequencing-based tests required, and also reduce invasive procedures and miscarriage rates. This test combination was modeled with final detection rates nearly as good as the integrated screen alone, but in the first trimester. However, sequencing-based testing alone still captured the most cases.
Whether sequencing-based screening is used as a replacement for traditional serum screening or as a follow-on test, there is a large impact on the number of miscarriages of euploid (no trisomy 21) miscarriages following an invasive procedure. With traditional screening in a high-risk population, about 20–30 normal fetuses are lost per 100,000 women screened; using sequencing-based testing the numbers drop to 2 or fewer. For low-risk women, 10–20 normal fetuses are lost per 100,000 women screened; with the use of sequencing-based testing, the numbers drop to 1 or none.
Another strategy, not shown in the decision model, is also possible: screening by sequencing-based testing without confirmatory testing. This would take advantage of the high detection rate and avoid the disadvantages of an invasive procedure and consequent risk of miscarriage. For this strategy, the most important information is the false-positive rate, which should ideally be zero. However, studies to date report rare but occasional false positives.
Authors’ Conclusions and Comment
This Assessment addressed the analytic and clinical validity, and clinical utility of nucleic acid sequencing-based testing primarily for Down syndrome (trisomy 21) compared to traditional screening procedures. Detection of Down syndrome cases is the original clinical reason for testing; standard screening tests may also detect trisomy 18 and 13. Thus, our report is primarily focused on results for trisomy 21, with discussion of results for trisomy 18 and 13.
There is little information on analytic validity. The available sequencing-based tests have not been submitted to the U.S. Food and Drug Administration (FDA) for regulatory review, and are offered as laboratory-developed tests subject only to laboratory operational oversight under CLIA. In recent years, recommendations for good laboratory practices for ensuring the quality of molecular genetic testing for heritable diseases and conditions under CLIA have been published. However, next-generation sequencing technology in general is new to the clinical laboratory, and regulatory and professional organizations are only beginning to address important issues of methods standardization.
Several studies of assay performance relative to the gold standard of karyotyping in high-risk populations were available. Some were multi-site studies that incorporated specimen collection, transport, and evaluation under conditions simulating real-world clinical testing. Review of study quality overall found low risk of bias except in the domain of patient selection. A majority of studies reported insufficient information on how patients were enrolled, and/or on reasons for exclusion prior to testing. Risk of bias in this domain was largely unclear due to lack of information. However, the impact on performance characteristics of the assay and ultimately on pregnancy outcomes is likely to be low, with one exception. The single study in an average screening population was judged to have a high risk of bias due to exclusions (some unavoidable) likely to affect case detection. In this study, cases were verified primarily by phenotype at birth from medical records, a poor standard compared to karyotyping.
In general, assays from all three companies1 currently offering fetal trisomy screening by sequencing DNA in maternal plasma show good clinical validity, with high sensitivity and specificity for Down syndrome (trisomy 21) and for trisomy 18. Few studies reported results for trisomy 13 and few cases were available in those that did, making it difficult to characterize overall performance for trisomy 13. All calculated negative predictive values for Down syndrome are near or at 100%, close to ideal for screening. Calculated positive predictive values vary considerably with risk of trisomy 21 in the tested population. Notably, however, false-positive rates were relatively invariant across a wide spectrum of T21 prevalence values. As more experience is gained with testing, it will be necessary to carefully document the false-positive rate for each assay.
Determination of clinical utility depends on a comparison with current screening practices and evaluation of impact on the outcomes of case detection, invasive confirmatory procedures required, and miscarriages resulting from invasive procedures. Actual comparative outcomes were not available, but instead were calculated from the summarized data on sequencing-based assay performance for trisomy 21, and published data on traditional screening performance, patient uptake of confirmatory testing, and miscarriage rates associated with invasive procedures to acquire confirmatory samples.
For each comparison and in each risk population, sequencing-based testing improved outcomes. As an example, if there are 4.25 million births in the U.S. per year and two-thirds of the population of (~average risk) pregnant women accept screening, then of about 2.8 million screened with the integrated screen, 74,434 will have an invasive procedure (assuming 50% uptake after a positive screening test and a recommendation for confirmation), 370 will have a miscarriage, of which 342 will be normal (non-trisomy 21) fetuses, and 3,417 of 3,559 Down syndrome cases will be detected. Using sequencing-based testing instead of traditional screening reduces the number of invasive procedures to 6,378 and the number of miscarriages to 32 (after amniocentesis; 14 normal) or 70 (after CVS; 31 normal), while increasing the cases detected to 3,531 of 3,559 possible, using conservative estimates. False negatives are conservatively estimated at 266 of 2.8 million women screened (0.01%) and may be lower, indicating that invasive testing after a negative result would have more risk than benefit.
Another testing strategy is to add sequencing-based testing only after a positive first trimester traditional combined screen, which in the prior scenario would decrease invasive procedures further to 3,103, miscarriages would decrease to 34 after CVS (only 1 normal), but only 2,990 of 3,559 cases of Down syndrome/trisomy 21 would be detected. Thus, while this strategy has the lowest rate of miscarriages of which only one represents a normal fetus, and the lowest rate of invasive procedures, it detects fewer cases than sequencing-based testing alone. Clearly, a strong advantage of using sequencing-based assays, either in place of traditional serum screening or as a follow-on assay, is that miscarriages of normal fetuses during confirmatory invasive procedures are considerably reduced.
These results are likely to apply to lower risk/prevalence populations because negative predictive value changes very little. Positive predictive value changes considerably, however, and confirmatory testing is strongly recommended for both low- and high-risk populations. Sequencing-based testing without confirmatory testing carries the risk of misidentifying normal pregnancies as positive for a trisomy syndrome due to the small but finite false-positive rate together with the low baseline prevalence of trisomy 21, 18, and 13 in all populations.
The decision model discussed in the preceding text was applied only to Down syndrome. However, based on the assay performance data for trisomy 18, which is very similar to that for trisomy 21, it is likely the outcome trends would be similar for trisomy 18. The data for trisomy 13 are too sparse for specific conclusions, but there is no biologic reason to suggest outcomes would be different.
While sequencing-based testing appears most effective as a replacement for traditional screening, it is not a replacement for ultrasound testing. The first trimester ultrasound scan that confirms gestational age and determines whether the pregnancy is multiple also provides necessary information for sequencing-based testing. The ultrasound exam that details the fetal anatomy in the second trimester is important for fetal risk assessment and may detect indications of chromosomal abnormalities in addition to those tested by currently available sequencing-based tests. Sequencing-based testing is also not a replacement for second trimester maternal serum AFP screening for risk of neural tube defects. The replacement of current maternal serum screening with sequencing-based testing would likely be accompanied by operational changes in screening programs and procedures and the need for provider education.
Limitations of sequencing-based tests include an indeterminate test rate (due either to a low fetal DNA fraction in the maternal plasma sample, to a deliberately chosen “no call” zone, or to unexplained assay failure) that may be as low as 1% or as high as about 5%, depending on the assay. Fetal fraction is determined either in an initial, separate test or as a part of the trisomy test by all companies. For three companies, the value of the fetal fraction is a quality control criterion. Below an established cutoff value, the sample is not acceptable for reporting assay results. The highest indeterminate rates do not reflect poor assay technology, but rather a choice in the case of one company’s assay to increase the accuracy of positive results by assigning results near the assay cutoff to a “no call” category. Indeterminate results require follow-up, which could be repeat testing of a new sample, since fetal fraction increases with time during pregnancy. Currently, however, there are no data on repeat testing. In addition, repeat testing adds the delay of a new sample collection in addition to the assay turnaround time for results. Alternatively, patients with indeterminate results could elect to proceed directly to an invasive procedure and karyotyping.
Each assay is currently specific for certain aneuploidies, and expansion of services is likely in the near future. For example, two of four companies now report detection of sex chromosome aneuploidies and a third reports Y chromosome detection for prenatal sex determination (used in genetic counseling for X-linked disorders). Published data were too few to evaluate this indication for this Assessment.
Looking toward future developments, there are broader implications for sequencing-based evaluation of fetal DNA in maternal plasma. Currently available tests include RhD blood type, fetal sex determination (clinically useful if, for example, a woman is a carrier of an X-linked condition such that a male fetus would be at risk), and detection of the aneuploidies discussed in this Assessment. However, it may be possible to use the technology to detect microdeletions and single-gene disorders. Moreover, the feasibility of mapping an entire fetal genome using this technology has been demonstrated. In short, an excess of information may be possible. Thus, some have called for “standardized regulations and guidelines that can harness the potential benefits and minimize the risks of non-invasive prenatal testing.”
Based on the available evidence, the Blue Cross and Blue Shield Association Medical Advisory Panel (MAP) made the following judgments about whether nucleic acid sequencing-based testing of maternal plasma meets the Blue Cross and Blue Shield Association Technology Evaluation Center (TEC) criteria to detect trisomy 21 in women being screened for fetal trisomy syndromes.
1. The technology must have final approval from the appropriate governmental regulatory bodies.
None of the commercially available sequencing assays for trisomy 21 has been submitted to or reviewed by the U.S. Food and Drug Administration (FDA). Clinical laboratories may develop and validate tests in-house (laboratory-developed tests or LDTs; previously called “home-brew”) and market them as a laboratory service; LDTs must meet the general regulatory standards of the Clinical Laboratory Improvement Act (CLIA). Laboratories offering LDTs must be licensed by CLIA for high-complexity testing.
2. The scientific evidence must permit conclusions concerning the effect of the technology on health outcomes.
Eight studies reported on the performance of DNA sequencing-based trisomy 21 screening in singleton high-risk pregnancy populations with invasive confirmatory procedures planned or completed. A ninth study in an average-risk singleton pregnancy population primarily compared DNA sequencing-based testing to a less accurate standard, phenotype at birth. The results of these studies provided strong estimates of assay performance characteristics for trisomy 21. Results for assay performance characteristics compared to the gold standard of karyotyping along with already available evidence on the performance of standard screening panels and confirmatory testing allowed the construction of a simple decision model to compare the health outcomes of nucleic acid sequencing-based testing with standard testing for trisomy 21.
3. The technology must improve the net health outcome, and
4. The technology must be as beneficial as any established alternatives.
In a decision model, sequencing-based maternal plasma trisomy 21 testing reduced the number of invasive confirmatory procedures needed and consequent associated miscarriages, while improving the number of detected cases of trisomy 21, compared to standard screening procedures in either high- or average-risk populations of pregnant women.
5. The improvement must be attainable outside the investigational settings.
Four of 9 studies were conducted by third-party investigators at multiple clinical locations (13–60 sites) in the U.S. and other countries; all companies’ assays were represented and samples were sent to company laboratories for sequencing-based testing, as would occur for routine clinical test orders. Thus, the test performance leading to improved overall screening outcomes should be attainable outside the investigational settings.
Based on the above, nucleic acid sequencing-based testing of maternal plasma for trisomy 21 with confirmatory testing of positive results (as is expected to be performed in a real-world clinical setting) in both high-risk women and average-risk women being screened for trisomy 21 meets the TEC criteria.
1As this Assessment was in press, a fourth company’s test became clinically available. A supporting “proof of concept” (author’s words) paper was published, but did not use the final calculation algorithm, so the report was not included as evidence in this review.
Full studies are in PDF format. You will need Adobe Acrobat Reader to view all studies.
Download Adobe Acrobat Reader here.
abnormal; abnormalities; AFP; allelic; alpha fetoprotein; amniocentesis; amplified; aneuploidies; aneuploidy; anomalies; anomaly; antenatal; cell-free DNA; chorionic villus sampling; chromosomal; chromosome; chromosome selective sequencing; CLIA; clinical laboratory; confirmatory testing; congenital; CVS; cytotrophoblastic; decision model; direct karyotypic; DNA; Down; epigenetic; euploid; fetal; first trimester; genetic; guidelines; heritable diseases; human genome; invasive; karyotyping; mapping; massively parallel sequencing; maternal; microdeletion; miscarriage; molecular; MPS; nex-gen; next generation; nucleic acid; Patau; PCR; plasma; polymerase chain reaction; polymorphic loci; prenatal diagnosis; prevalence; screen; second trimester; sensitivity; sequencing; shotgun sequencing; specificity; syndrome; trisomy; whole genome;