gine non-invasive prenatal testing for aneuploidy

Ultrasound Obstet Gynecol 2013; 42: 1533Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/uog.12513

Non-invasive prenatal testing for aneuploidy: current statusand future prospects

P. BENN*, H. CUCKLE and E. PERGAMENT*Department of Genetics and Developmental Biology, University of Connecticut Health Center, Farmington, CT, USA; Department ofObstetrics and Gynecology, Columbia University Medical Center, New York, NY, USA; Northwestern Reproductive Genetics, Chicago,IL, USA

KEYWORDS: amniocentesis; aneuploidy; chorionic villus sampling; Down syndrome; fetal DNA; maternal plasma; screening;sequencing; trisomy

ABSTRACT

Non-invasive prenatal testing (NIPT) for aneuploidyusing cell-free DNA in maternal plasma is revolutionizingprenatal screening and diagnosis. We review NIPT in thecontext of established screening and invasive technologies,the range of cytogenetic abnormalities detectable, cost,counseling and ethical issues. Current NIPT approachesinvolve whole-genome sequencing, targeted sequencingand assessment of single nucleotide polymorphism (SNP)differences between mother and fetus. Clinical trialshave demonstrated the efficacy of NIPT for Down andEdwards syndromes, and possibly Patau syndrome, inhigh-risk women. Universal NIPT is not cost-effective,but using NIPT contingently in women found at moderateor high risk by conventional screening is cost-effective.Positive NIPT results must be confirmed using invasivetechniques. Established screening, fetal ultrasound andinvasive procedures with microarray testing allow thedetection of a broad range of additional abnormalitiesnot yet detectable by NIPT. NIPT approaches thattake advantage of SNP information potentially allow theidentification of parent of origin for imbalances, triploidy,uniparental disomy and consanguinity, and separateevaluation of dizygotic twins. Fetal fraction enrichment,improved sequencing and selected analysis of the mostinformative sequences should result in tests for additionalchromosomal abnormalities. Providing adequate prenatalcounseling poses a substantial challenge given the broadrange of prenatal testing options now available. Copyright 2013 ISUOG. Published by John Wiley & Sons Ltd.

INTRODUCTION

The past quarter century has been witness to a series ofremarkable advances in the screening of pregnancies foraneuploidy, particularly in the identification of Downsyndrome (trisomy 21). During the 1970s and early

Correspondence to: Dr P. Benn, Department of Genetics and Developmental Biology, University of Connecticut Health Center, Farmington,CT, USA (e-mail: [email protected])

Accepted: 14 May 2013

1980s, advanced maternal age, defined in most localitiesas over 35 years, was the only means by which thegeneral population was assessed as to risk of a fetalchromosomal abnormality. Fewer than one third ofDown syndrome pregnancies were diagnosed prenatallyand of those undergoing invasive prenatal diagnosis onlyabout 2% had fetal karyotype abnormalities1, a figurecomparable to the 0.51% chance of procedure-relatedfetal loss associated with amniocentesis or chorionic villussampling (CVS)2. In the late 1980s and early 1990s, theintroduction of second-trimester maternal serum markers,in the form of double, triple and quad marker testing,improved significantly the screening performance for ane-uploidy. The proportion of Down syndrome pregnanciesdiagnosed more than doubled and a chromosomal abnor-mality was found in as many as 4% of those designatedas screen-positive3. In the late 1990s and early 2000s,aneuploidy screening shifted to the first trimester withthe combined test, which uses ultrasound measurementof nuchal translucency thickness (NT) together withmaternal serum concentration of placental proteinshuman chorionic gonadotropin (hCG) (free , intactor total) and pregnancy-associated plasma protein-A(PAPP-A). Currently available screening protocols alsoincorporate additional ultrasound markers and sequentialscreening using two blood samples, one in the firstand one in the second trimester, with or without NT.Consequently, screening performance has improved suchthat more than nine-tenths of Down syndrome cases canbe diagnosed prenatally4 and the yield from invasivetesting has risen to about 6%5.

Recently, analysis of cell-free (cf)DNA in maternalblood for non-invasive prenatal testing (NIPT) has beenshown to be highly accurate in the detection of commonfetal autosomal trisomies. In 1997, Lo et al.6 first reportedtheir seminal discovery that plasma from pregnant womencontained cfDNA, including a fraction designated fetal,although this is thought to be placental in origin, resulting

Copyright 2013 ISUOG. Published by John Wiley & Sons Ltd. REVIEW

16 Benn et al.

from apoptotic trophoblasts. This and other early studies7

suggested that the fetal fraction was only 36%, butmore recent studies have found that it may be closerto 1020%8. Fetal cfDNA can be detected as early as4 weeks gestation9 and exceeds 4% of all the cfDNA innearly all women from 10 weeks onward. The typicalsize of the cfDNA fragments is approximately 150basepairs10,11 and, importantly, the entire fetal genomeis represented. It has been shown that the half-life offetal cfDNA is very short12; fetal fragments are no longerdetectable very soon after birth13,14. There is thereforeno serious concern that a prenatal cfDNA test could beconfounded by a prior pregnancy in multigravid women.Maternal plasma cfRNA, unlike RNA extracted directlyfrom cells, appears to be relatively stable and can alsopotentially be used in screening1517.

The commercial introduction of NIPT raises severalconcerns. One major issue is the optimal integration ofNIPT into current screening practice18. Should NIPT andconventional screening tests be treated as related butindependent measures when assessing aneuploidy risk orshould they be complementary to one another? How canhealthcare providers be educated and prospective patientscounseled about all the prenatal testing options nowavailable?

In this Review we provide the information needed byclinicians and public health providers when consideringto whom NIPT will be offered and regarding theconsequences of doing so, practical suggestions on how toimplement this new and powerful technology into routineclinical practice, and some indications of how we expectthis testing to expand. We focus primarily on the detectionof fetal chromosomal abnormalities, recognizing that thetechnology will eventually be used for testing for a broadrange of other genetic disorders.

ESTABLISHED SCREENING MODALITIES

In the current context it is important to distinguish pre-natal diagnosis of aneuploidy from antenatal screening. Adiagnostic test performed on chorionic villi, amniotic fluidor fetal blood needs to have very few false negatives (aneu-ploid pregnancies misdiagnosed as euploid) and false pos-itives (euploid pregnancies misdiagnosed as aneuploid),since the result will inform the decision as to whether toterminate the pregnancy. In contrast, antenatal screeningdoes not aim to be definitive; rather, it is designed to iden-tify women who are at sufficiently high risk of commonaneuploidies as to warrant invasive prenatal diagnosis.Since the invasive diagnostic procedures, mainly CVS andamniocentesis, are hazardous and expensive, only a rela-tively small group of women are identified as being at highrisk. Established screening programs determine the risk ofDown syndrome, many also including Edwards syndrome(trisomy 18) and some including Patau syndrome (trisomy13) and Turner syndrome (45,X). The model-predictedaneuploidy detection rates (DR) for fixed false-positiverates (FPR) of established prenatal screening protocolshas been reviewed elsewhere4.

Among those undergoing invasive prenatal diagnosisbecause of a high risk of Down and Edwards syndromes,a large proportion have a different chromosomal abnor-mality. Alamillo et al.19 reported on 10 years experienceof a combined test, giving risks of Down syndromeand Edwards or Patau syndromes. The 97 cases withchromosomal abnormalities included Down syndrome(48%), Edwards syndrome (16%), Patau syndrome (6%),Turner syndrome (9%), other sex chromosomal aneu-ploidies (4%), other aneuploidies including mosaics (9%)and chromosomal rearrangements (6%). The Californiastate-wide screening program20 used a second-trimesterquad test for Down and Edwards syndromes and detecteda total of 1316 chromosomal abnormalities. Excludingbalanced translocations, these were Down syndrome(48%), Edwards syndrome (14%), Patau syndrome(2%), Turner syndrome (8%), triploidy (2%), Klinefeltersyndrome (47,XXY) (1%) and other abnormalities(25%). Some of these were chance findings but manywere related to maternal age or abnormal marker profiles.

Conventional aneuploidy screening modalities alsoneed to be considered in the context of the substantialnumber of non-chromosomal fetal abnormalities andpregnancy complications that may be identified as a resultof this testing. Particularly important is the role of first-trimester ultrasound, which has the potential to identifya non-chromosomal abnormality in approximately 1%of cases5.

ESTABLISHED INVASIVE PRENATALDIAGNOSIS

Conventional cytogenetics

Chromosomal analysis of cultured chorionic villi andamniotic fluid cells can identify fetal aneuploidy,structural rearrangements, such as translocations andinversions, and relatively large duplications and deletions(generally exceeding 5 Mb in size)21. Chromosomalanalysis of amniotic fluid cells is considered to be thegold standard in prenatal testing because error rates areexceedingly low, probably less than 0.010.02%, andmostly appear to be due to maternal cell contamination,laboratory error or typographic mistakes22. Error rateswith CVS are higher than those seen with amniocentesis,and these can be due to confined placental mosaicism(CPM), maternal cell contamination or lower resolutionbanding of chromosomes23. In a small proportion ofcases in which CVS karyotyping is carried out thereare ambiguous results which may require resampling oramniocentesis for clarification.

Molecular cytogenetics

Because of the time taken to obtain a complete chromo-somal analysis, the occasional finding of a karyotype ofminimal or uncertain significance and cost considerations,some centers perform a rapid aneuploidy test (RAT)using molecular genetic techniques as an adjunct to or

Copyright 2013 ISUOG. Published by John Wiley & Sons Ltd. Ultrasound Obstet Gynecol 2013; 42: 1533.

NIPT for aneuploidy 17

replacement for karyotyping. Methods include the useof fluorescence in-situ hybridization (FISH)24,25, quanti-tative fluorescence polymerase chain reaction (QF-PCR)testing26 and multiplex ligation-dependent amplification(MLPA)27. These tests are generally designed to test only arestricted range of cytogenetic abnormalities28. Althoughproven to be very accurate, confirmatory testing usingconventional cytogenetics has been recommended29.

In contrast, relative to conventional karyotyping, othermolecular approaches can now improve substantially onthe detection of clinically significant genetic imbalances.In high-risk pregnancies with a normal chromosomalconstitution, the increased resolution of array compara-tive genome hybridization (aCGH) coupled with singlenucleotide polymorphism (SNP) array has the potential toidentify submicroscopic copy number variations (CNVs),specifically microdeletions and microduplications, as wellas regions of homozygosity.

When compared to conventional chromosomal analy-sis, aCGH and SNP array are purported to have significantadvantages in terms of: (1) faster turnaround time;(2) higher throughput while being less labor-intensive;(3) higher resolution independent of the ability of thecells to grow and/or generate good metaphase spreads;(4) amenability to automation and quality control; and,(5) direct mapping of aberrations to the genome sequence.Faster turnaround time is possible because arrays canbe performed on DNA obtained from uncultured villiand amniotic fluid cells, thus avoiding the need for tis-sue culture and the associated time delay required forconventional chromosomal analysis.

A National Institute of Health (NICHD) multicenter,prospective, blinded study30 evaluated the applicationof microarrays in over 4000 high-risk pregnanciesundergoing prenatal invasive testing largely because ofadvanced maternal age, a positive first-trimester screenor the presence of an ultrasound anomaly. The use ofmicroarrays resulted in the identification of clinicallyrelevant CNVs that were unrecognizable by conventionalkaryotyping in 2.5% of all cases. In those cases referredfor testing because of an ultrasound abnormality andshowing a normal karyotype, 6% were found to havea clinically relevant CNV. These studies were performedon low-resolution aCGH platforms; newer arrays havemuch higher resolution. The use of SNP arrays hasexpanded the ability to detect triploidy and significantregions of homozygosity, thereby detecting uniparentaldisomy and potentially identifying disorders associatedwith consanguinity. The NIH multicenter study alsorevealed the limitations of all aCGH and SNP arrayplatforms, the principal limitation being the identificationof variants of uncertain significance (VOUS).

Hazards of invasive testing

The principal hazard of amniocentesis is miscarriage, butthe excess risk associated with the procedure is difficultto quantify precisely. Some 34% of mid-trimesterpregnancies will miscarry without amniocentesis and in

a particular case of fetal loss following the procedure itis possible only rarely to attribute directly the adverseoutcome to the procedure; cases of amnionitis or chronicamniotic fluid leakage would be attributable but theseare relatively rare consequences. In a randomized trial ofamniocentesis at a single center31, the fetal loss rate was0.8% higher in the amniocentesis group compared withcontrols. There have been no similar randomized trialsfor CVS, but a number of studies have compared latefirst-trimester CVS with second-trimester amniocentesis.A Cochrane Review32 showed that, when performed by askilled operator, the fetal loss rates of the two procedureswere comparable. An updated review2 concluded that theexcess miscarriage rate for either procedure is between0.5% and 1%. An initial concern that CVS may causefetal limb reduction defects has not been sustained, asthe reported excesses occurred when CVS was performedearly in gestation, before 10 weeks32.

NIPT METHODS

The fact that fetal cfDNA and cfRNA are present onlyas minority components of the total cell-free nucleic acidsin maternal plasma specimens has posed a significanttechnical obstacle in the development of NIPT. Anumber of approaches have been suggested that allowenrichment of the fetal component. These include takingadvantage of size differences between fetal and maternalDNA fragments11, using formaldehyde in the samples33,taking advantage of the nucleosome binding propertyof the DNA34 and immunoprecipitation of hypo- orhypermethylated DNA35, but none is currently beingused in clinical practice. Even without enrichment anumber of technical approaches have now been provento be effective. Each approach relies on the extraordinaryadvances in molecular biology and sequencing achievedin recent years. We briefly review the major methodsbelow, without discussing in detail all of the possiblerefinements and other developments that are currentlyunder consideration.

Shotgun massively parallel sequencing (s-MPS)

This approach relies on the identification and countingof large numbers of the DNA fragments in the plasmaspecimen. Using massively parallel sequencing (MPS),millions of both fetal and maternal DNA fragments canbe sequenced simultaneously and, since the entire humangenome sequence is known, each piece that maps to adiscrete locus can be assigned to the chromosome fromwhich it came36,37. If fetal aneuploidy is present, thereshould be a relative excess or deficit for the chromosomein question. The difference in counts will be small; forexample, if fetal trisomy 21 is present and the fetalfraction is 20%, the relative excess in chromosome21 DNA fragments will be (0.8 2) + (0.2 3) = 2.2compared with the situation for a euploid fetus(0.8 2) + (0.2 2) = 2, i.e. a relative increase in numberof chromosome 21 counts of only 10%. To be able to


18 Benn et al.

detect reliably these differences, large numbers of countsare necessary, the fetal fraction needs to be appreciableand the counts need to be compared with the expectedcounts for euploid cases. The latter can be achieved eitherby normalizing counts for the chromosome of interestagainst other chromosomes that are expected to bedisomic within the same test run36 or by comparing thefraction of counts assigned to a particular chromosomeagainst the fractions seen for a set of known euploidcases38. Results can be expressed as a z-score whichcan be converted into a probability that a specificchromosome result departs from the normal diploidsituation. Although it is possible to convert the z-scoreto a patient-specific risk, major commercial providersof this testing currently choose to present their resultsonly in terms of positive or negative based on z-scoresexceeding predefined thresholds. Another variation of thestatistical analysis calculates two Student t-test statistics,one based on the hypothesis that the result is from aeuploid fetus and the other that it is from an aneuploidfetus and then calculates a further test statistic, L,which is the logarithmic likelihood ratio39. Refinementsin the statistical analysis also include adjustment forguanine-cytosine (GC) base content of the sequences40,41.

The approach is referred to as shotgun because it relieson sequencing and counting all informative chromosomeregions. Sequences which do not map to a unique locusare uninformative. In some studies, only those fragmentsthat have sequences identical to the reference genomesequence are accepted in the analysis, while in othersone or two mismatches in the sequence will be accepted.Because of costs, it is common to run multiple samplessimultaneously (multiplexing) by adding a bar-codesequence to each fragment that will allow identificationof the sample origin. The number of cases run togethercan be variable and is an important consideration becausehigh-level multiplexing places a limitation on the totalnumber of sequence reads that are carried out on eachcase (i.e. the depth of sequencing). The number offetal sequence reads will also vary according to thefetal fraction and therefore a minimum fetal fraction,typically 4%, is currently required. Most studies haveused Illumina sequence analyzers but there are also studiesusing Solid42,43 and Helicos44 platforms.

Targeted massively parallel sequencing (t-MPS)

It is also possible to amplify selectively only those chro-mosomal regions that are of interest (e.g. chromosomes21, 18 and 13) and then to assess whether there is adeparture from euploidy based on the relative number ofDNA fragment counts for this subset of chromosomes45.This targeted approach has the advantage of requiringconsiderably less sequencing and thereby reduces costs.The major commercial provider of this approach includesan adjustment to allow for the variability in the propor-tion of DNA that is fetal and then combines the results ofthe laboratory test with maternal age to provide a patient-specific risk for Down, Edwards and Patau syndromes45.

In theory, results could also be combined with other fac-tors, for example results of other screening tests or priorhistory of an aneuploid pregnancy.

Single nucleotide polymorphism (SNP)-basedapproaches

The feasibility of a non-invasive aneuploidy test that takesadvantage of DNA polymorphisms was first demonstratedby Dhallan et al.46. In their approach, the paternal,maternal and fetal SNPs present in blood (paternal),buffy coat (maternal) and maternal plasma (maternaland fetal) were evaluated as bands on sequencing gels.Quantification of band intensities of the uniquely inheritedpaternal SNPs in the plasma allowed an estimation ofthe fetal fraction. Comparison of the maternal plusfetal bands with the unique fetal band intensity alsoallowed estimation of the fetal chromosome 21 dosage.A similar approach was used by Ghanta et al.47, whoidentified combinations of paired SNPs to focus on highlyinformative regions and then used cycling temperaturecapillary electrophoresis for evaluating fetal fraction anddosage. The method was thought to be useful in situationsin which the fetal fraction was as low as 2%.

A more complex approach that relies on polymor-phisms has been proposed by Zimmermann et al.48. Theapproach involves a multiplex amplification of 11 000SNP sequences in a single PCR reaction carried out onthe plasma DNA followed by sequencing. Each product isevaluated based on the hypothesis that the fetus is mono-somic, disomic or trisomic. After considering the positionsof the SNPs on the chromosomes and the possibilitythat there may have been recombination, a maximumlikelihood is calculated that the fetus is either normal,aneuploid (chromosome 21, 18, 13 or sex chromosome)or triploid, or that uniparental disomy is present.

In theory, with SNPs there should be advantages overmethods based on total sequence counting. SNPs canprovide information about parent of origin of aneuploidy,recombination and inheritance of mutations. On theother hand, SNPs account for only 1.6% of the humangenome and therefore enrichment for fetal DNA, deepersequencing or higher levels of high-fidelity amplificationmay be required to identify unambiguously affectedpregnancies with small imbalances49. When few loci areused in the analysis, greater attention needs to be paid tothe exclusion of regions in which there are rare benignCNVs. The SNP approach could result in the inadvertentdetection of non-paternity or perhaps consanguinity. Inthe case of assisted reproduction with an egg from anunrelated donor, SNP-based approaches would need tobe modified to take into consideration the presence ofadditional maternally derived fetal alleles not present inthe surrogate mother.

The laboratory methods and data analyses used forSNP-based NIPT are substantially different from thoseused in shotgun and targeted counting approaches. In thisReview we therefore consider the emerging clinical trialdata for SNP-based testing separately.



Methylated DNA-based approaches

This approach takes advantage of the fact that therewill be epigenetic differences throughout the genomewhich will result in the differential expression of genes,depending on the tissue type50. These differences can beassociated with either hyper- or hypomethylation of thegenes. Some genes present in placental cells (specifically,trophoblasts) will therefore differ from maternal cellsin the methylation pattern. Methylated DNA can bechemically modified, thus allowing its characterizationseparate from the unmethylated DNA51. Alternatively,the hypermethylated DNA can be analyzed separatelyfollowing restriction enzyme digest of the hypomethylatedDNA52. If, additionally, there are polymorphic differencesin the fetal alleles, disturbance from a 1:1 ratio can beused to identify trisomy. This has been demonstrated fordetection of Edwards syndrome but the approach has notbeen tested in clinical trials51. Differentially methylatedregions may be relatively common53 so the method mayalso be applied to detect other aneuploidies.

It is also possible to use immunoprecipitation withantibodies specific to 5-methylcytidine to enrich for themethylated DNA. For example, the identification of a setof specific DNA sequences on chromosome 21 that arehypermethylated in placental cells allows the selectiveenrichment of fetal cfDNA from maternal plasma35.Following quantitative real-time PCR, there may bemeasurable differences in the amount of chromosome 21-derived DNA when fetal trisomy 21 is present, comparedwith in euploid fetuses. One proof-of-principle study hasdemonstrated the efficacy of such an approach to NIPT54.A follow-up study using a modified set of differentiallymethylated chromosome 21 DNA sequences also showedencouraging results55. This approach is currently thesubject of further clinical trials. If an approach suchas this can be fully developed, it could be substantiallycheaper than sequencing-based methods.

Digital PCR

Another way in which the costs of sequencing can beavoided is through the use of digital PCR. This technologyis based on the dilution of test materials such thatindividual DNA fragments of interest (e.g. chromosome21) can be amplified separately and counted5658.Theoretical estimates of the number of amplificationevents have been computed59 and this approach couldbecome particularly attractive if reliable fetal cfDNAenrichment methods were developed.

RNA-based testing

Identification of genes that are expressed by trophoblastsbut not maternal cells should result in the presence ofthe corresponding cfRNA species in maternal plasma,free of contaminating maternal RNA with a similarsequence. If the cfRNA also carries polymorphisms thatdistinguish the fetal alleles, the presence of aneuploidy

could be deduced from departures from a normal 1:1ratio in the relative amount of the inherited alleles60. Thisapproach to NIPT showed some success in the initialproof-of-principle studies, but unpublished clinical trialdata were disappointing. Nevertheless, with the discoveryof additional RNAs that can be analyzed, this approachmay yet play an important role in NIPT61.

NIPT IN CLINICAL TRIALS:NON-MOSAIC DOWN, EDWARDS ANDPATAU SYNDROMES

Proof-of-principle studies demonstrated the feasibility ofNIPT using small series of maternal plasma samplesfrom aneuploid and euploid pregnancies3638,4046,6264.However, generally, the samples were not tested blindto the outcome of pregnancy, ascertainment criteria werenot fully described and some samples were obtained afteran invasive diagnostic procedure which could have ledto increased transfer of fetal DNA into the maternalcirculation. There have now also been several large clinicaltrials of MPS which have overcome these limitations andpotential biases39,6575. These have established NIPT as apotentially highly effective screening test but not as a testthat could replace current invasive prenatal diagnosis. Wetherefore prefer the term NIPT or cfDNA screening ratherthan non-invasive prenatal diagnosis, which is misleading.

High-risk studies

There have been seven cfDNA studies involvingshotgun or targeted sequencing in women who hadinvasive prenatal diagnosis because of high risk for:Down syndrome65,67, Down syndrome or anotheraneuploidy66,68,69 or any disorder70,71. In one study, ascreen-positive screening test was the only indicationfor high risk69, whilst in the four other studies in whichthe reason for invasive testing was high risk for Downsyndrome or another aneuploidy, the indications weremixed. The reason for invasive testing was a screen-positive screening test in 77%65, 43%67, 30%66 and22%68 of cases; advanced maternal age in 68%66, 37%(only indication in these cases)68 and 33%67 of cases;ultrasound in 13%66, 11%67 and 7%68 of cases; familyhistory in 5%66, 3%67 and 3%68 of cases; and multiplereasons in 22%68 and 9%67 of cases. Five studies includedresults on both Down and Edwards syndromes6872 andthree also included Patau syndrome68,72,75.

The study design details for these trials on high-riskwomen are summarized in Table 1. One study designatedas unclassified results with z-scores of 2.54.0 forautosomal chromosomes and excluded them from theDR and FPR calculations68. Doing so is not valid76,particularly when no protocol was suggested for theunclassified group, and in this Review we have reclassifiedthem as negative. However, the unclassified group had afive-fold increased likelihood of aneuploidy (3.5% (5/144)of trisomies unclassified compared with 0.7% (9/1358) ofcontrols) and it could be argued that they should in fact


20 Benn et al.

Table 1 Eleven cell-free DNA studies of shotgun or targeted sequencing: methodological differences

Trial Design Average GA (weeks) Method Algorithm Cut-off*

High riskChiu et al.65 Prospective and casecontrol 13 Shotgun z-score 3.0Ehrich et al.66 Casecontrol 16 Shotgun z-score 2.5Palomaki et al.67,72 Casecontrol 15 Shotgun z-score 3.0Bianchi et al.68 Casecontrol 15 Shotgun z-score 4.0Sparks et al.70 Casecontrol 18 Targeted Risk 1%Ashoor et al.69,75 Casecontrol 13 Targeted Risk 1%Norton et al.71 Prospective 16 Targeted Risk 1%

Other than high riskLau et al.73 Prospective 14 Shotgun z-score? ?Nicolaides et al.74 Retrospective 12 Targeted Risk 1%Dan et al.39 Prospective 20 Shotgun z-score & L-score 2.5 or 1.0Gil et al.78 Prospective 10 Targeted Risk 1%

*For autosomal chromosomes. Includes maternal and gestational age. Results with scores 2.54.0 were unclassified. Positive if z1. GA, gestational age at testing.

be reclassified as positive, so those rates are also given inthis Review.

Table 2 summarizes the Down and Edwards syndromesresults for these seven cfDNA studies carried out in high-risk pregnancies. In total these studies compared 594samples from pregnancies with Down syndrome with5745 that did not, and 193 samples from pregnancieswith Edwards syndrome with 5459 that did not, althoughin both cases some had other aneuploidies. The resultsfrom the various studies, notwithstanding methodologicaldifferences, are consistent with each other and theircombined data represent the best estimates of DRand FPR. For Down syndrome, the overall DR of99.3% has a 95% CI of 98.299.8% and FPR of0.16% (95% CI, 0.080.31%). Even with the mostextreme CI values this performance exceeds, by far,anything achievable by current Down syndrome screeningprotocols. However, this performance falls short of thatfor current diagnostic tests. Furthermore, reclassifying aspositive the unclassified results in the study of Bianchiet al.68 would have increased the overall DR slightly to99.5%, but almost doubled the FPR to 0.26%. Hence,we can conclude that based on present evidence cfDNAtesting is not a diagnostic test of Down syndrome butrather is a very good secondary screening test. A womanwith a positive cfDNA test will be at high risk of

Down syndrome but depending on her pretest risk itmay not be extremely high. This can be illustrated fromthe combined data in Table 3, by estimating a positivelikelihood ratio (LR) using the DR divided by the FPR,i.e. 634. For example, a woman with a pre-test risk of1 in 270 following a combined test would have a post-test odds of 634:269 or a risk of 2 in 3. This is notsufficiently high as to warrant termination of pregnancywithout a confirmatory invasive diagnostic test. Similarly,the negative LR, estimated from the false-negative ratedivided by the true negative rate, is 1/148. A womanwith a pre-NIPT risk of 1 in 10 would have a 1 in 1333risk after a negative test, which might be insufficientlyreassuring for some women. These LRs are illustrativeand cannot be used to counsel individual women becausesome of the studies had already incorporated prior riskbased on maternal age into their algorithms. However,it is reasonable to conclude that using NIPT following ascreen-positive test for any of the established screeningprotocols followed by confirmatory invasive diagnostictesting if the NIPT result is positive will probably hardlyaffect the overall screening DR but will reduce the FPRabout 300-fold.

For Edwards syndrome, the overall DR is 97.4%(95% CI, 93.799.0%) and the FPR 0.15% (95% CI,0.070.31%), with positive LR of 665 and negative

Table 2 Down syndrome and Edwards syndrome results in seven cell-free DNA studies carried out in high-risk pregnancies

Down syndrome Edwards syndrome

Trial DR (n (%)) FPR (n (%)) DR (n (%)) FPR (n (%))

Chiu et al.65* 86/86 (100) 3/146 (2.1) Ehrich et al.66 39/39 (100) 1/410 (0.24) Palomaki et al.67,72 209/212 (98.6) 3/1471 (0.20) 59/59 (100) 5/1688 (0.30)Bianchi et al.68 89/90 (98.9) 0/410 (0.00) 35/38 (92.1) 0/463 (0.00)Sparks et al.70 36/36 (100) 1/123 (0.81) 8/8 (100) 1/123 (0.81)Ashoor et al.69 50/50 (100) 0/297 (0.00) 49/50 (98.0) 0/297 (0.00)Norton et al.71 81/81 (100) 1/2888 (0.03) 37/38 (97.4) 2/2888 (0.06)

Total 590/594 (99.3) 9/5745 (0.16) 188/193 (97.4) 8/5459 (0.15)

*2-plex and 8-plex data were reported but only the more favorable 2-plex results are included here. DR, detection rate; FPR, false-positiverate.



Table 3 Failure rates in 11 studies of cell-free DNA testing*

Trial Failure rate (n (%)) Reasons for failure

Chiu et al.65 11/764 (1.4) Low total DNA (n= 2), low DNA library concentration (n= 8), low matched DNA sequencereads (n= 1)

Ehrich et al.66 18/467 (3.8) Low % fetal DNA (< 4%) (n= 7), low total DNA (< 556 copies) (n= 7), low DNA libraryconcentration (< 32.3 nM) (n= 15), low number unique DNA sequence counts (< 3million) (n =11); some failed more than one criteria

Palomaki et al.67 13/1696 (0.8) Low % fetal DNA (< 4%) (n= 6), other QC parameters (n= 7): low DNA libraryconcentration (< 25 nM) and low matched DNA sequence reads (< 12.5 million)

Bianchi et al.68 16/532 (3.0) No fetal DNASparks et al.70 8/338 (2.4) Low % fetal DNA (< 3%), low DNA sequence counts, evidence from SNPs of non-singleton

pregnancyAshoor et al.69 3/400 (0.8) Amplification and sequencingNorton et al.71 148/3228 (4.6) Low % fetal DNA (< 4%) (n= 57), assay failure (n= 91): inability to measure % fetal, high

variation in DNA counts and failed sequencingLau et al.73 0/567 (0.0) Nicolaides et al.74 100/2049 (4.9) Low % fetal DNA (< 4%) (n= 46), assay failure (n= 54)Dan et al.39 79/11 184 (0.7) Quality of separation, extraction, sample preparation and sequencing: low peak DNA library

size, low library concentration (< 30 nM) and low matched DNA sequence reads (< 2million)

Gil et al.78 40/997 (4.0) Low % fetal DNA (< 4%) (n= 23), assay failure (n= 17)*Excluding tests rejected because of inadequate sample quality.

LR 1/38. Including the unclassified results in the studyof Bianchi et al.68 as positive would have increasedthe DR to 98.4% and the FPR to 0.20%. Since acfDNA test is unlikely to be aimed at detecting Edwardssyndrome alone, these false-positives should be regardedas additional to those generated by cfDNA testing forDown syndrome. There are no data with which to assesswhether there is any tendency for those testing falselypositive for one type of aneuploidy to be falsely positivefor another. Assuming these are uncorrelated, the FPR forboth aneuploidies together is 0.30%, or 0.46% if thoseunclassified by Bianchi et al.68 are included.

Combining the results from the three studies whichincluded Patau syndrome68,72,75 gives a DR of 30/38(78.9%; 95% CI, 65.991.9%) and FPR of 17/4112(0.41%; 95% CI, 0.220.61%), with a positive LR of191 and a negative LR of 1/4.7. Including the unclassifiedresults as positive, the DR increased to 80.0% with noincrease in the FPR. On the basis of these results, an MPStest for all three trisomies would have a FPR of 0.70.9%,depending on the inclusion of the unclassified results.

Initial prospective clinical experience has been reportedfor a commercial laboratory, Verinata Inc, offeringtests to high-risk women77. While follow-up informationwas incomplete, among 5974 samples there were falsepositives for Down, Edwards and Patau syndromes andfalse negatives for Down and Edwards syndromes, withrates compatible with those seen in the trial data.

Non-high risk

There have been four NIPT studies involving shotgunor targeted sequencing in which the subjects were notspecifically selected because they had invasive prenataldiagnosis. Their methods are summarized in Table 1.

Lau et al.73 reported prospective cfDNA screeningresults of 567 patients, 49% of whom were tested

at 1213 weeks gestation. These women comprised179 (32%) who were screen-positive by conventionaltesting, 194 (34%) who were screen-negative or awaitingresults and 194 (34%) who were unscreened. There wasinsufficient follow-up to report DRs reliably, but eightcases of Down syndrome and one of Edwards syndromewere detected and there were no false positives. Seven ofthe detected cases were among the screen-positive women(1 in 25) and two, both Down syndrome, were amongthe remainder (1 in 194).

Nicolaides et al.74 carried out a retrospective studyusing stored plasma samples from women who had acombined test at 1113 weeks gestation. After excludingnon-viable pregnancies, those terminated for reasonsother than Down or Edwards syndromes and cases lostto follow-up, 1949 were successfully tested for cfDNA,including eight with Down syndrome and two withEdwards syndrome. All aneuploidies were detected and,while there were no false-positive Down syndrome casesamong 1939 tested, there were two (0.05%) false-positiveEdwards syndrome cases among 1937 tested. Invasive pre-natal diagnosis had been carried out in 4.3% (86/2049) ofthose selected, including all of the cases with aneuploidy.

A large prospective study of cfDNA screening wascarried out in China39. A total of 11 105 pregnancieswere tested successfully, including 4522 (41%) thatwere screen-positive following conventional screening and2770 (25%) with risk factors such as advanced maternalage, ultrasound abnormalities and family history. Allof the known Down syndrome (n= 140) and Edwardssyndrome (n= 42) cases were detected by the cfDNAtest. However, approximately one third of the 10 915with negative test results were lost to follow-up; 2818(26%) had invasive testing, a further 70 (< 1%) wereterminated, and follow-up was available for 4524 (41%).The protocol included an insurance scheme whereby thefamily would be compensated in the event of unpredicted


22 Benn et al.

delivery of either a Down or an Edwards syndromebaby. This is likely to be a considerable incentive toreport false-negative results so the lack of follow-up maynot be a major problem. There were two proven false-positive cfDNA results, one each for Down and Edwardssyndromes. However, a further eight positive cases didnot have a karyotype following fetal loss or terminationand therefore the possibility of additional false positivescannot be excluded.

Gil et al.78 carried out a prospective study of cfDNAscreening at 10 weeks gestation in 997 women withmedian age 37 years; combined tests were also carriedout. A total of 15 cases of aneuploidy were detected bycfDNA and confirmed by invasive testing (10 cases ofDown syndrome, four of Edwards syndrome and oneof Patau syndrome) and there was one false positive(for trisomy 18). One case was positive for trisomy 21but miscarried before invasive confirmation and no falsenegatives had emerged at the time of writing.

These studies provide some, albeit limited, evidenceto suggest that cfDNA screening may be as effectivein the general population as it is in those alreadyscheduled for invasive testing: the FPR is not dissimilarto that in the high-risk studies; all aneuploidies appearto have been detected, although most of them were inpregnancies which were already candidates for invasivetesting. Moreover, among high-risk women, performancedoes not appear to be related to the indication for invasivetesting. Palomaki et al.67 found that the DRs and FPRswere very similar according to indication: screen-positive99% (96/97) and 0.2% (1/637), respectively; advancedmaternal age or family history 100% (24/24) and 0.3%(2/587); ultrasound 100% (51/51) and 0% (0/130);and multiple reasons 95% (37/39) and 0% (0/112).Nevertheless, until much larger general population studiesare published, the limited direct evidence of efficacy needsto be supplemented by indirect evidence. The separationin the distributions of z-scores or risks between aneuploidand euploid pregnancies is necessarily dependent on thefetal fraction and it is therefore important to considerwhether this might be lower in low-risk women.

Several covariables for fetal fraction have beeninvestigated, including indication for invasive prenataldiagnosis, maternal age, screening marker level andestimated aneuploidy risk. In one published series ofpregnancies at high risk of Down syndrome or otheraneuploidy67, there was no significant difference in meanz-score, in either Down syndrome or euploid pregnancies,according to indication or correlated with maternal age,but there was a significant negative correlation withmaternal weight. The latter effect can be explained bythere being a fixed mass of fetal DNA, from the placenta,mixing with an increasing mass of maternal DNA; suchan effect is also seen in conventional maternal serummarkers. In another published series of pregnancies athigh risk of Down syndrome69, a multivariate regressionanalysis was carried out which confirmed that maternalweight was a covariable but maternal age was not79. Inaddition it showed that there was a significant positive

correlation between fetal fraction and both PAPP-A andfree -hCG, which might be related to placental volume,although there was no association with gestational age ineither study. Since in Down syndrome pregnancies PAPP-A is reduced on average and free -hCG is increased, thecorrelations with fetal fraction may cancel each other out.In Edwards syndrome pregnancies, both serum analytemarkers are reduced, consistent with reduced placentalvolume80. There could therefore be a somewhat increasedchance for cfDNA test failure in affected pregnancies.On the other hand, those cases with higher PAPP-A and-hCG which might not be detected by the combined testmay show higher fetal fractions and these may therefore bemore amenable to detection by a cfDNA-based NIPT. In afurther published series of those having invasive prenataldiagnosis due to high risk generally71, the mean fetalfraction was computed for deciles of maternal age, NTand aneuploidy risk81. There was no significant differencein means between the highest and lowest deciles. Thehighest decile for risk comprised 100 women with riskgreater than 1 in 10 and a mean fetal fraction of 11.4%compared with 135 women in the lowest decile with riskless than 1 in 6500 and a mean fetal fraction of 10.8%.

Failed results

Several of the MPS studies discussed above specified thenumber of pregnancies not considered for cfDNA testingbecause of sample quality. Overall, among 13 260 eligiblepregnancies there were 814 (6.1%) untested. This rateranged from 0.8% to 9.9% between the studies and mayreflect the stringent standards required when carryingout a research project. Of more concern is the 2.0% ofpregnancies (436/22 222) in which a cfDNA test wasperformed but failed because of insufficient fetal DNAor other technical reasons (Table 3). The failure ratemay be even higher in those resampled after an initialfailure: it was 32% (13/40) in one study78. Qualitycontrol criteria varied between the studies but in thosethat included a minimum acceptable percentage of fetalDNA, about half the failures were due to this not beingmet. In the prospective laboratory experience reportedby Futch et al.77, a result was not possible in 2.4% ofcases. This is relevant to comparisons with conventionalscreening, which rarely reports failure to obtain a result.

Three prospective studies of women not specificallyselected because they had invasive prenatal diagnosisprovide information on the extent to which a successfultest result was achieved following an initial failure. In onestudy73, among 567 successful cfDNA tests, four (0.7%)had required a repeat blood sample and no other tests hadfailed during that period (T.K. Lau, pers. comm.). Thetime from the initial blood draw to the cfDNA report wasgreater than 14 days for 16 (2.8%) tests, including thoserequiring a repeat sample. In another study39, among11 105 tests, 97 (0.9%) needed resampling. In this studyas a whole, the turnaround time was within 10 workingdays for 95% of pregnancies and within 15 for 99%;those requiring resampling took a further 1015 days to



produce a report. In the third study82, a request for apatient redraw was made in 1.9% of referrals and amongthose who did undergo a redraw there was sufficient fetalcfDNA in the second sample in 56% of cases. The successrate was strongly dependent on maternal weight.

SNP studies

Following the initial proof-of-concept study using poly-morphisms and maximum LRs for test interpretation48,the results of one small clinical trial have been published83

and additional information has been presented at scien-tific meetings in the form of an abstract84 and a poster85.The study of Nicolaides et al.83 involved 242 singletonpregnancies with results available for 229 (95%). All 32aneuploid and 210 euploid cases were identified correctly.The abstract84 provided information on 407 pregnan-cies including 44 with aneuploidies and the results were100% correct among samples that passed quality control.In addition to the maternal plasma, paternal blood orbuccal mucosa samples were genotyped. Whilst this wasnot an absolute requirement (paternally inherited allelesin the fetus can be deduced), the failure rate was lowerwhen a paternal sample was available. In the poster85,there were 763 samples, of which 45 (5.9%) failed to passquality control, and among the remaining 717 (including47 Down, 15 Edwards, 7 Patau and 12 Turner syndromes)samples all except one result, for Turner syndrome, wascorrect.

Multiple pregnancies

There is a potential problem with using cfDNA toscreen for aneuploidy in multiple pregnancies. As withbiochemical screening, in twins discordant for trisomy theexcess in fragments from a specific chromosome in theaffected twin may be masked by the normal proportionof those fragments in the euploid cotwin. This would alsobe the case for higher order multiple pregnancies. Thisproblem might be overcome by using SNPs to measurethe variations in the fetal fraction between genomicregions86,87. The method allows determination of zygosityand, in dizygous twins, examination of the distribution offetal fragments for each fetus.

At present, only two small multiple-pregnancy cfDNAseries have been published, using the same method asfor singletons88,89. The first series was derived from oneof the high-risk studies67 and included three discordanttwins (two with Down and one Patau syndrome), fiveconcordant Down syndrome twins, 17 euploid twins andtwo euploid triplets88; there were no false-positive or false-negative cfDNA test results. The second series was fromone of the studies of women not selected because theywere at high risk73 and included only one discordanttwin with Down syndrome and 11 euploid twins89;all pregnancies were correctly classified by the cfDNAtest.

In a singleton pregnancy that had had a vanishing twin,it is theoretically possible that apoptosis of cells from

the fetoplacental remains of the non-viable fetus couldinterfere with the cfDNA result; there is a relatively highchance that the fetal loss would have been associatedwith aneuploidy and this could lead to a false-positiveresult77.

COST OF NIPT

Currently, the cost of a cfDNA test is high, ranging fromabout $800 to $2000 in the USA, and from $500 to$1500 elsewhere. Whilst this may not necessarily deterindividuals from being tested, it is likely to determinewhether public health planners provide the testing andwill be the basis for whether health insurance schemesreimburse those tested. We have published a sensitivityanalysis90 in which the factors contributing to theseoverall costs are elucidated for four cfDNA screeningpolicies: (1) cfDNA screening for those screen-positive oncombined test; (2) universal cfDNA screening, replacingall current screening modalities; (3) contingent cfDNAscreening, for the 1020% of women with the highestrisk based on a combined test; (4) cfDNA screening forwomen of advanced maternal age and younger womenwith a positive combined test90.

Restricting cfDNA testing to screen-positive cases leadsto a small reduction in detection, a massive reduction infalse positives and is the least expensive way in whichto utilize cfDNA testing. With this policy the averagecost per Down syndrome birth avoided is dependent onlyon the difference in unit cost of a cfDNA test and ofinvasive prenatal diagnosis. For a combined test withFPR of 3%, the average cost would be reduced by usingsecondary cfDNA testing, if the unit cost was less than orequal to three-quarters that of invasive testing. Even at ahigher unit cost of cfDNA, the average cost per fetal lossprevented would be relatively low.

For other policies in which the aim of cfDNA testingis to increase detection, the most important public healthconsideration will be not the total or average cost butthe marginal cost compared with the existing screeningprovision. The marginal cost is the average cost per Downsyndrome birth avoided or loss prevented over and beyondthat which would have been avoided by the existingservice. This is shown in Table 4 for fixed unit costs forthe combined test and invasive testing and with variableunit costs of cfDNA. The table also shows screeningperformance.

Universal cfDNA screening yields both a very high DRand a small FPR but the marginal cost is very high,exceeding the lifetime cost to the system of a Downsyndrome birth9194. With universal screening the costper fetal loss prevented when replacing conventionalscreening is also very high. Contingent protocols, in whichall women receive the combined test and the 1020% withthe highest risk receive cfDNA testing, are efficient andcan reduce massively the marginal cost compared withuniversal screening. The average cost of preventing a fetalloss is also several times lower than that for universalcfDNA testing. A policy of cfDNA testing only for older


24 Benn et al.

Table 4 Four cell-free (cf)DNA policies*: performance, marginal cost per Down syndrome (DS) birth avoided and average cost per fetal lossprevented, according to unit cost of cfDNA testing

Policy Performance

Marginal cost per DS birthavoided ($ million) forunit cost cfDNA of:

Average cost per fetal lossprevented ($ million) forunit cost cfDNA of:

Combined cfDNA DR (%) FPR (%) OAPR $1500 $1000 $500 $1500 $1000 $500

None All 99.3 0.16 1:1 5.84 3.63 1.42 9.47 5.89 2.31All Contingent:

10% 90.4 0.016 7:1 1.11 0.65 0.19 0.82 0.48 0.1415% 92.8 0.024 5:1 1.39 0.86 0.32 1.33 0.82 0.3020% 94.4 0.032 4:1 1.67 1.05 0.44 1.85 1.16 0.48

< 35 years 35 years 85.0 2.1 1:19 2.90 1.73 0.57 2.35 1.41 0.46< 35 years 35 years or positive

on combined test84.7 0.017 6:1 3.69 2.02 0.37 0.81 0.44 0.08

*Three reported in a modeling analysis90 and the fourth, cfDNA restricted to older women, using the same model. Combined test,false-positive rate (FPR) = 3%, unit cost = $150; invasive prenatal diagnosis, unit cost = $1000, iatrogenic fetal loss rate = 0.5%. DR,detection rate; OAPR, odds of being affected given a positive result.

women was not substantially more effective than wasthe combined test and is more costly than contingentscreening. A hybrid policy of offering cfDNA testing toall those of advanced maternal age plus all combined testscreen-positive women was better as it reduced the FPRmassively but it was less effective and more expensivethan was a contingent testing policy.

The detection rate for a contingent policy can beincreased by including additional serum and ultrasoundmarkers. Nicolaides et al.95 suggest markers which arealso of value in screening for pre-eclampsia and fetalcardiac abnormalities. Cost-benefit analyses for suchvariants have not yet been carried out.

Two other published studies have estimated costs whencfDNA is applied to women with positive conventionalscreening results, using the estimated NIPT DR and FPRfrom their own single study. Palomaki et al.67 consideredtotal costs following screening in 100 000 screen-positivewomen, with an average risk of 1 in 33 for a Downsyndrome fetus. They concluded that if the unit cost ofcfDNA was less than 96% of the unit cost of invasivetesting, there could be a net saving (4% of costs needto be assigned to test failures and confirmatory invasivetesting following cfDNA positive results). The secondstudy also considered 100 000 screen-positive women,modeled as a mixture of combined test or second-trimesterserum screening96. They estimated total costs includingboth the conventional screen and subsequent testing andconcluded that with a unit cost of cfDNA $1200, CVS$1700 and amniocentesis $1400, there would be a 1%cost reduction. One other study has estimated costsfor cfDNA screening in older women and in youngerwomen with screen-positive conventional screening tests,using DR and FPR estimates from the literature97. Theauthors claimed that this policy, in the USA, would havelower total cost compared with conventional screening.However, this conclusion appears to have been reachedfollowing incorrect assumptions about the DR and FPR inyounger women and a very low uptake of invasive testingfor women screen-positive by conventional screeningcompared with cfDNA screening90.

FUTURE PROSPECTS: NIPT FORADDITIONAL CYTOGENETICABNORMALITIES

Other autosomal abnormalities

With the widely anticipated rapid improvements insequencing, it may soon be possible to screen forother autosomal trisomies in situations in which thefetal fraction is very low, thereby allowing detection ofaneuploidy even earlier in pregnancy. Early pregnancyis associated with very high levels of chromosomalabnormality and although nearly all abnormal caseswill spontaneously abort, early identification could bebeneficial to women. First-trimester autosomal trisomycan involve any chromosome, but some occur much morefrequently than others22. Detection of these through NIPTwill be facilitated by the larger size of the additionalchromosome in the non-viable trisomies. Early studiessuggested that there was greater variability in the fragmentcounts for some of the larger chromosomes36, but thiscould be overcome by GC correction and selective use ofthe most consistent regions of chromosomes37,62.

Screening for a large number of potential abnormalitiestogether does require very high specificity for eachcomponent so that the aggregate FPR for all chromosomescombined is low.

Sex-chromosome abnormality

Establishing NIPT for sex-chromosome imbalances isdesirable because some specific sex-chromosomal abnor-malities are associated with a sufficiently severe phenotypeas to fully justify prenatal diagnosis: for example, manycases of Turner syndrome, most of which will not sur-vive to full term, gain of multiple X chromosomes andsome unbalanced structural rearrangements of the X-chromosome. Collectively, unbalanced sex-chromosomalabnormalities constitute over a third of the unbalancedkaryotypes seen in amniocenteses performed in womenunder 35 years of age and nearly a quarter of those inolder women98. There is mosaicism in approximately one



third of sex chromosomal aneuploidies detected throughamniocentesis98.

One commercial laboratory that provides NIPT,Sequenom Inc, was routinely providing information onthe presence or absence of Y-chromosome DNA in theirreporting. Based on a series of 1639 samples, the DRfor a Y-chromosome was 828/833 (99.4%), and theaccuracy for the prediction of a female fetus was 803/806(99.6%)99. More recently, they have been providingNIPT for Turner syndrome, Klinefelter syndrome,47,XXX and 47,XYY based on a study by Mazloomet al.100, with DRs of 83% (25/30), 85% (11/13), 83%(5/6) and 75% (3/4), respectively, for each of the theseabnormalities. The FPR for Turner syndrome was 0.2%(4/1945) and there was one other false positive (47,XXX)in the series100. Of the normal cases in which gender wasinterpreted, it was inconsistent with the karyotype resultin 0.7% (13/1814)100. Bianchi et al.68 developed a com-plex algorithm for evaluating sex-chromosome-derivedcfDNA in plasma. Their study derived z-scores for theX-chromosome (zX) and Y-chromosome (zY). A result wasclassified as positive for Turner syndrome if zX

26 Benn et al.

mesenchyme, but the reverse situation can also be present.An abnormal cell line can also be present in both direct andlong-term cultures but be undetectable in the fetus. Longerterm cell culture is usually considered to be a more reliableindicator of the true karyotype of the fetus comparedwith direct preparations. Overall, approximately 1% ofall CVS specimens will show a second cell line that isnot detectable in the fetus, a situation termed confinedplacental mosaicism (CPM)113. This may sometimesresult in spontaneous loss, low birth weight or pretermdelivery, particularly when it involves both trophoblasticand mesenchymal cell lineages114 and/or when it involvesspecific karyotype abnormalities such as trisomy 16115.However, in most pregnancies in which CPM is detectedthe pregnancy outcome is normal116.

The frequency and significance of CPM is relevant toNIPT because cfDNA is believed to originate primarilyfrom non-viable trophoblasts. CPM could result in eitherfalse-positive or false-negative NIPT results. It has beensuggested that viable pregnancies with Edwards or Patausyndrome may have a substantial euploid cell line inthe trophoblasts117,118 and therefore NIPT could fail todetect some viable trisomic cases in which there is asubstantial population of abnormal cells in the fetus butnot in the placenta. This would lead to an apparent false-negative result by NIPT. Conversely, the trophoblastsmay be mostly abnormal and the fetal cells normal,leading to an apparent false-positive result. Moreover, ifchromosomally abnormal trophoblasts were more likelyto undergo apoptosis, NIPT could preferentially identifycases in which the proportion of abnormal cells in theplacenta is low.

A number of cases have already been described whichillustrate how differences in the distribution of normaland abnormal cells may lead to test result discrepanciesbetween NIPT and karyotyping. Faas et al.43 described theidentification of Turner syndrome based on NIPT whenthe abnormal cell line was confirmed by cytogenetic anal-ysis of short-term CVS cultures but the abnormality wasnot seen in the long-term CVS cultures or in the newbornblood. Choi et al.119 described a case involving possibletrisomy 22 detected by cfDNA testing. The trisomy 22cell line was confirmed by analyzing the term placentabut the abnormality was not detected in a blood specimenfrom the dysmorphic baby. Hall et al.120 describe acase with a positive Patau syndrome cfDNA test, mosaic47,XY,+13/46,XY result for CVS, normal result for amni-otic fluid cells, normal result for cord blood, but trisomy13 mosaicism confirmed in the placenta. These examplescan each be interpreted as CPM although the presence oflow-level true fetal mosaicism cannot be excluded.

Early trials on chromosomal analysis following CVSprovide an indication as to how often CPM might lead toan apparent false-positive or false-negative result. Becausemilder abnormalities such as sex chromosomal aneuploidyor mosaicism involving abnormalities not typically seenin livebirths may not have been fully pursued in follow-up studies, this analysis needs to be confined to Down,Edwards and Patau syndrome. Ledbetter et al.23 noted six

cases (four involving chromosome 13, one chromosome18 and one chromosome 21) interpreted as a false-positiveCPM result and one false-negative result (chromosome18) among 6560 cases that had analysis of a direct CVSpreparation. Similarly, Smith et al.121 reported nine cases(two involving chromosome 13, four chromosome 18 andthree chromosome 21) interpreted as a false-positive CPMresult and five false-negative results (two chromosome 18,three chromosome 21) in approximately 18 195 casesthat had direct preparation analysis. Combining thesedata, for chromosome 21, 18 or 13, the incidence offalse positives was 1 in 1650 and the incidence of falsenegatives was 1 in 4126. These estimates are crude becausethe number of cells analyzed on direct CVS preparationis often small, some women receiving an abnormal resultterminated their pregnancy without confirmation, andlow-level true fetal mosaicism may have gone undetected.It should be noted that most cases of CPM do notinvolve chromosomes 21, 18 and 13. Therefore, as NIPTtests are expanded to include additional chromosomes orchromosomal regions, the issue of erroneous results dueto CPM is likely to become increasingly important.

For those cases that do show true fetal mosaicism, thephenotype can be highly variable but will generally be lesssevere than that present with the non-mosaic abnormalkaryotype22. Although the proportion of abnormal cellsin amniotic fluid or chorionic villi can be a poorguide to phenotype in individual cases, in general, ahigh proportion of abnormal cells is associated withadverse pregnancy outcome122. NIPT tests that onlypresent a result categorically as either positive ornegative obviously cannot provide any informationabout the proportion of abnormal cells or clinical severity.Furthermore, since some of the clinical trials havespecifically excluded mosaic cases and optimized theirz-score cut-off on the basis of non-mosaic cases, testing ispresumably biased towards not detecting mosaic cases. Itis not yet clear whether patients with z-scores close to thecut-off should be offered invasive testing to help rule outmosaic trisomy. To present this problem appropriatelyto clinicians, and also to allow better integration of theresults with other screening and diagnostic techniques,we recommend that NIPT reports should be formattedso that they provide patient-specific risks for each non-mosaic trisomy, the z-score and the fetal fraction, ratherthan just a categorical positive or negative finding.

The possibility of maternal mosaicism also needs to beconsidered. This includes the rare situation in which theabnormal maternal cfDNA is derived from a malignantcell population123. Maternal mosaicism could potentiallybe recognized using SNP-based NIPT methodology.

Translocations

Unbalanced translocations, in theory, should be identi-fiable in NIPT as a result of the partial trisomy andmonosomy that is present. In a retrospective analysis ofsix plasma specimens from pregnant women with knownfetal imbalances, Srinivasan et al.112 were able to identify



the imbalances in maternal plasma in all six cases. In twoof their cases the karyotype showed additional material ofunknown origin and the NIPT analysis was able to providea more precise characterization of the extra chromosomalmaterial. The ability to detect routinely smaller imbal-ances would require deeper sequencing and improvementsin bioinformatic analysis of sequencing data41,124126.For SNP-based approaches to NIPT, detection of smallimbalances requires identification of sufficient informativepolymorphisms in the region of interest.

Conventional chromosomal analysis does allow detec-tion of most apparently balanced translocations and largeinversions. Prenatally identified de-novo rearrangementsare associated with an increased risk for abnormality127.It would therefore be desirable to detect these non-invasively. Paired-end sequencing of DNA fragments thatspan a chromosomal breakpoint can identify at least sometranslocations through the analysis of cellular DNA. Forexample, Schluth-Bolard et al.128 used a whole-genomepaired-end protocol to identify the specific sequence dis-ruptions in four cases with apparently balanced transloca-tions and abnormal phenotypes. Breakpoints were mostlyin repeat sequences and most breakpoints were associatedwith gain or loss of some basepairs. Talkowski et al.129

used sequencing to refine breakpoints and establish adiagnosis of CHARGE syndrome based on disruption ofthe CHD7 gene by analyzing amniotic fluid cells. Thesestudies involved the sequencing of larger pieces of DNAand were carried out in cases in which the karyotypeinformation was used to help identify candidate DNAfragment sequences that spanned the breakpoints.

Although the routine detection of de-novo rearrange-ments using cfDNA, without knowledge of where to look,would not seem to be technically achievable at this time,it is potentially possible. The challenge is similar to thedetection of other mutations130132, but is made sim-pler with the presence of two related reciprocal events.By sequencing cfDNA it should be possible to constructthe haplotype sequence in which one end matches per-fectly to a chromosome, the other end matches perfectlya second chromosomal location and there are similarhybrid sequences present that correspond to the recipro-cal translocation product(s). Improved sequencing fidelity,deeper sequencing and the development of advanced bio-metric algorithms that specifically look for the hybridsequences would be needed.

Inherited rearrangements are generally consideredbenign, but their detection through karyotyping follow-ing CVS or amniocentesis is considered advantageousbecause it identifies a risk factor for future pregnancies.When a balanced translocation is already known to besegregating in a family and the phase of SNPs on thechromosomes in the region of interest can be established,the presence or absence of the breakpoint can be inferredfrom the SNPs in the cfDNA in a plasma specimen. In thecase of a paternally inherited translocation, the presenceof the translocation chromosomal SNPs and absence ofthe normal homolog SNPs identifies the translocation inthe conceptus. In maternal inheritance, detection of the

translocation relies on the identification of a relative excessof the translocation chromosomal SNPs and relative defi-ciency of the normal homolog SNPs in the plasma sample.

Triploidy

Triploid pregnancies can arise as a result of anadditional set of maternal (digynic) or paternal (diandric)chromosomes. Most recognized cases show a 69,XXXor 69,XXY karyotype; 69,XYY forms are thought toundergo early first-trimester miscarriage133. Most casesof triploidy will be identified by ultrasound abnormalityand characteristic first- or second trimester-maternalserum marker levels134136. It is exceptional for thesepregnancies to survive into the third trimester.

Since all chromosomes are proportionately representedin a 69,XXX pregnancy, NIPT based on the countingof all chromosomal sequences would yield a normalresult. Pregnancies with either a 69,XXY or a 69,XYYkaryotype might be identifiable through unexpected sexchromosomal sequence ratios. Digynic triploidy mightbe particularly difficult to detect by NIPT because theplacenta is very small and the fetal fraction may thereforebe reduced137. Consistent with this, Bianchi et al.68

reported that of nine cases of fetal 69,XXX triploidy, therewas insufficient fetal cfDNA for analysis in three (33%).

Using the approach that takes into consideration theSNPs present in the cfDNA, diandric triploidy could beidentified through the presence of two different paternallyinherited SNPs at some loci. Digynic cases with sufficientfetal cfDNA for analysis might be recognized throughthe quantitative excess of maternally inherited fetal DNAfragments relative to paternally inherited fragments.

CNVs

Small duplications and deletions have the potential to bedetected through NIPT. Examples that have been detectedinclude a 4-Mb deletion on chromosome 12138, and twocases in which 3-Mb 22q11.2 deletions were present139.In a series of 11 prenatal plasma specimens, Srinivasanet al.112 identified small gains or losses not reported inthe clinical karyotypes in eight cases, including four withtwo such variations. However, it is not clear whether theywere true or false positives.

A variety of strategies have been proposed to detect thesmall deletions, insertions, inversions, and duplicationsthat are common in the human genome140. Much ofthe analysis is contingent on the ability to identifydepartures in the location and orientation of sequencedDNA fragments from where they are expected to be foundin reference mapped genomes. The ability to do this typeof analysis in a non-invasive test is a long-term goal inmolecular cytogenetics.

Overview of future testing

Currently, only methods based on MPS have been thesubject of clinical trials and can be viewed as being


28 Benn et al.

suitable for clinical use. An optimal NIPT for fetalaneuploidy will be accurate, simple, cheap, applicableearly in pregnancy and fully compatible with existingprenatal screening methods so that risks developed usingthe different approaches can be combined. Many of themethods under consideration can be refined to allow forthe detection of other chromosomal imbalances. MPSrefinements that will aid this process include selectionof optimal sequences for analysis, adjustments for basecomposition and the use of sequencing platforms thatprovide higher read accuracy.

Sequencing also provides opportunities to detect thepresence or absence of single gene disorders130. Themethods can even be applied to the construction of anentire fetal genotype131,132. On the other hand, methodsthat do not require sequence information from the parentsor which do not generate unwanted additional genotypicstatus of the parents (for example resulting in informationabout paternity or other disease risks) should also beconsidered advantageous. The optimal technologies mayultimately be those which are most amenable to answeringonly those clinical questions that are being asked, ratherthan those which provide the most information.

COUNSELING

The internationally accepted approach to dealing withpatients diverse ethical, religious and cultural values is toprovide individual non-directive counseling, testing andother clinical information at each step in the screening anddiagnosis process. This model, which maximizes patientautonomy in reproductive decision making, has beenadopted by diverse populations throughout the world.

Counseling of women who are considering NIPT ischallenging. The range of cytogenetic abnormalities cur-rently detectable through cfDNA testing is smaller thanthat detectable by conventional karyotyping and substan-tially less than that achievable through microarray testing.In the absence of public health policies that define testingstrategies, each high-risk woman who previously wouldhave made a decision about amniocentesis or CVS andkaryotyping, will now need to choose between no testing,NIPT, invasive testing with conventional cytogenetics orinvasive testing with microarray testing. These women willneed to be counseled carefully regarding the complex ben-efits, hazards and trade-offs associated with each option.

In part, these decisions will be determined by theindication for testing and perceived risk, but they will alsobe determined by the available alternative testing services.For example, in a situation in which routine invasivetesting is based on RAT for a limited set of disorders, thechoice of NIPT that detects the same range of aneuploidiesmight seem preferable. On the other hand, offering arraytechnology may provide much more valuable informationconcerning fetal wellbeing than can current NIPT. Thisbenefit has to be weighed against the risk of losing anormal pregnancy following an invasive procedure or theheightened anxiety caused by the presence of a CNV ofunknown significance, poorly defined penetrance and/or

variable expressivity. In addition, the testing could exposeconsanguinity or non-paternity. Given these realities, thechallenge now facing those providing pretest counselingcannot be understated. The National Society of GeneticCounselors recognizes that due to limited resources,pretest counseling cannot always be provided by agenetic counselor and this information will therefore alsoneed to be communicated by other qualified healthcareproviders141.

The availability of professional guidelines, suchthose from the American College of Obstetricians andGynecologists, together with the Society for Maternal-Fetal Medicine142, the Society of Obstetricians andGynaecologists of Canada143, the National Society ofGenetic Counselors144, the American College of MedicalGenetics and Genomics145 and the International Societyfor Prenatal Diagnosis (ISPD)146, can greatly facilitatethe process by clarifying which processes are scientificallyvalid and clinically acceptable standards of care.

ETHICAL ISSUES

The promise of NIPT is that it will reduce the number ofwomen with an indication for invasive testing and onlythose women who are at very high risk for defined disor-ders such as Down syndrome will then need to deal withthe ethical challenges associated with procedure-relatedloss and pregnancy termination. However, the ease ofproviding NIPT will potentially result in substantiallymore prenatal diagnoses and terminations of affectedpregnancies18. The issue is confounded by an inability toprovide adequate counseling for all women who might beconsidering NIPT147. There may also be a shift in patientsand/or healthcare providers views on prenatal diagnosis,with a greater expectation for pregnancy terminationas the usual response to an abnormal diagnosis. Thishas been referred to as normalization of testing andtermination148. Potential consequences could include asharply reduced incidence of Down syndrome neonates,greater emphasis on prevention and less on support andaccommodation of their disabilities, and altered attitudesto individuals with Down syndrome and their parentswho chose not to test or terminate an affected pregnancy.

The currently available NIPT for aneuploidy is theconsequence of a massive investment by private companiesand venture capital, and so far is only provided by private,for-profit companies. Aspects of the testing are the subjectof contested patents and other proprietary ownership.Claims of exclusive intellectual property ownership caninflate costs, restrict access to test protocols and limitfurther test development. Although there is currentlyno direct-to-consumer testing, there are press releases,patient information materials and extensive marketing tohealthcare providers, many of whom may have a poorunderstanding of the true performance of the tests.

In the future, NIPT may also be encouraged by govern-ments or insurance carriers who view genetic disorders asan economic burden149. The concern is that these financialpressures will be the dominating factor in determining



who has access to testing, without adequately consideringmedical need. Changes in financial support for testing,counseling and clinical services can also indirectlycompromise individual patient autonomy when decidingwhether to continue or terminate an affected pregnancy.

It is also worth noting that the early history of NIPTshows how easily test quality can be compromised whendevelopment and introduction is left entirely to unregu-lated free market forces150,151. ISPD, in its latest PositionStatement on NIPT has strongly emphasized the impor-tance of adoption of laboratory standards for NIPT146.

DISCLOSURES

H. Cuckle is a consultant to PerkinElmer Inc, AriosaDiagnostics Inc, Natera Inc, and director of Genome Ltd;E. Pergament is a consultant to PerkinElmer Inc andNatera Inc, and director of Northwestern ReproductiveGenetics Inc.

REFERENCES

1. Ferguson-Smith MA, Yates JRW. Maternal age specificrates for chromosome aberrations and factors influencingthem: report of a collaborative European study on 52965amniocenteses. Prenat Diagn 1984; 4 Spec No: 544.

2. Tabor A, Alfirevic Z. Update on procedure-related risks forprenatal diagnosis techniques. Fetal Diagn Ther 2010; 27:17.

3. Benn PA, Egan JF, Fang M, Smith-Bindman R. Changes in theutilization of prenatal diagnosis. Obstet Gynecol 2004; 103:12551260.

4. Cuckle H, Benn P. Multianalyte maternal serum screeningfor chromosomal defects. In Genetic Disorders and the Fetus:Diagnosis, Prevention and Treatment (6th edn), Milunsky A,Milunsky JM (eds). Wiley-Blackwell, Chichester, UK, 2010;771818.

5. Syngelaki A, Chelemen T, Dagklis T, Allan L, NicolaidesKH. Challenges in the diagnosis of fetal non-chromosomalabnormalities at 1113 weeks. Prenat Diagn 2011; 31:90102.

6. Lo YMD, Corbetta N, Chamberlain PF, Rai V, Sargent IL,Redman CWG, Wainscoat JS. Presence of fetal DNA inmaternal plasma and serum. Lancet 1997; 350: 485487.

7. Lo YMD, Tein MS, Lau TK, Haines CJ, Leung TN, PoonPM, Wainscoat JS, Johnson PJ, Chang AM, Hjelm NM.Quantitative analysis of fetal DNA in maternal plasma andserum: implications for noninvasive prenatal diagnosis. Am JHum Genet 1998; 62: 768775.

8. Lun FM, Chiu RW, Allen Chan KC, Yeung Leung T, Kin LauT, Dennis Lo YM. Microfluidics digital PCR reveals a higherthan expected fraction of fetal DNA in maternal plasma. ClinChem 2008; 54: 16641672.

9. Illanes S, Denbow M, Kailasam C, Finning K, Soothill PW.Early detection of cell-free fetal DNA in maternal plasma.Early Hum Dev 2007; 83: 563566.

10. Chan KC, Zhang J, Hui AB, Wong N, Lau TK, Leung TN, LoKW, Huang DW, Lo YM. Size distributions of maternal andfetal DNA in maternal plasma. Clin Chem 2004; 50: 8892.

11. Li Y, Zimmermann B, Rusterholz C, Kang A, HolzgreveW, Hahn S. Size separation of circulatory DNA in maternalplasma permits ready detection of fetal DNA polymorphisms.Clin Chem 2004; 50: 10021011.

12. Lo YMD, Zhang J, Leung TN, Lau TK, Chang AM, HjelmNM. Rapid clearance of fetal DNA from maternal plasma.Am J Hum Genet 1999; 64: 218224.

13. Smid M, Galbiati S, Vassallo A, Gambini D, Ferrari A, VioraE, Pagliano M, Restagno G, Ferrari M, Cremonesi L. Noevidence of fetal DNA persistence in maternal plasma afterpregnancy. Hum Genet 2003; 112: 617618.

14. Hui L, Vaughan JI, Nelson M. Effect of labor on postpartumclearance of cell-free fetal DNA from the maternal circulation.Prenat Diagn 2008; 28: 304308.

15. Poon LL, Leung TN, Lau TK, Lo YM. Presence of fetal RNAin maternal plasma. Clin Chem 2000; 46: 18321834.

16. Ng EK, Tsui NB, Lam NY, Chiu RW, Yu SC, Wong SC, LoES, Rainer TH, Johnson PJ, Lo YM. Presence of filterableand nonfilterable mRNA in the plasma of cancer patients andhealthy individuals. Clin Chem 2002; 48: 12121217.

17. Ng EK, Tsui NB, Lau TK, Leung TN, Chiu RW, PanesarNS, Lit LC, Chan KW, Lo YM. mRNA of placental origin isreadily detectable in maternal plasma. Proc Natl Acad Sci U SA 2003; 100: 47484753.

18. Norton ME, Rose NC, Benn P. Noninvasive prenatal testingfor fetal aneuploidy: clinical assessment and a plea forrestraint. Obstet Gynecol 2013; 121: 847850.

19. Alamillo CML, Krantz D, Evans M, Fiddler M, PergamentE. Nearly a third of abnormalities found after first-trimesterscreening are different than expected: 10-year experience froma single center. Prenat Diagn 2013; 33: 251256.

20. Kazerouni NN, Currier RJ, Flessel M, Goldman S, HenniganC, Hodgkinson C, Lorey F, Malm L, Tempelis C, Roberson M.Detection rate of quadruple-marker screening determined byclinical follow-up and registry data in the statewide Californiaprogram, July 2007 to February 2009. Prenat Diagn 2011;31: 901906.

21. Hulten MA, Dhanjal S, Pertl B. Rapid and simple prenataldiagnosis of common chromosome disorders: advantages anddisadvantages of the molecular methods FISH and QF-PCR.Reproduction 2003; 126: 279297.

22. Benn PA. Prenatal diagnosis of chromosome abnormalitiesthrough amniocentesis. In Genetic Disorders and the Fetus:Diagnosis, Prevention and Treatment (6th edn), Milunsky A,Milunsky JM (eds). Wiley-Blackwell: Chichester, UK, 2010:771818.

23. Ledbetter DH, Zachary JM, Simpson JL, Golbus MS,Pergament E, Jackson L, Mahoney MJ, Desnick RJ, SchulmanJ, Copeland KL, Verlinsky Y, Yang-Feng T, Schonberg SA,Babu A, Tharapel A, Dorfmann A, Lubs HA, Rhoads GG,Fowler SE, De La Cruz F. Cytogenetic results from theU.S. Collaborative Study on CVS. Prenat Diagn 1992; 12:317345.

24. Evans MI, Henry GP, Miller WA, Bui TH, Snidjers RJ,Wapner RJ, Miny P, Johnson MP, Peakman D, Johnson A,Nicolaides K, Holzgreve W, Ebrahim SA, Babu R, JacksonL. International, collaborative assessment of 146,000 prenatalkaryotypes: expected limitations if only chromosome-specificprobes and fluorescent in-situ hybridization are used. HumReprod 1999; 14: 12131216.

25. Vialard F, Simoni G, Aboura A, De Toffol S, Molina GomesD, Marcato L, Serero S, Clement P, Bouhanna P, Rouleau E,Grimi B, Selva J, Gaetani E, Maggi F, Joseph A, BenzackenB, Grati FR. Prenatal BACs-on-Beads: a new technology forrapid detection of aneuploidies and microdeletions in prenataldiagnosis. Prenat Diagn 2011; 31: 500508.

26. Mann K, Hills A, Donaghue C, Thomas H, Ogilvie CM.Quantitative fluorescence PCR analysis of> 40,000 prenatalsamples for the rapid diagnosis of trisomies 13, 18 and 21 andmonosomy X. Prenat Diagn 2012; 32: 11971204.

27. Gerdes T, Kirchhoff M, Lind AM, Vestergaard Larsen G,Kjaergaard S. Multiplex ligation-dependent probe amplifica-tion (MLPA) in prenatal diagnosis-experience of a large seriesof rapid testing for aneuploidy of chromosomes 13, 18, 21, X,and Y. Prenat Diagn 2008; 28: 11191125.

28. Caine A, Maltby AE, Parkin CA, Waters JJ, Crolla JA;UK Association of Clinical Cytogeneticists (ACC). Prenataldetection of Downs syndrome by rapid aneuploidy testing for


30 Benn et al.

chromosomes 13, 18, and 21 by FISH or PCR without a fullkaryotype: a cytogenetic risk assessment. Lancet 2005; 366:123128.

29. Test and Technology Transfer Committee. Technical andclinical assessment of fluorescence in situ hybridization: AnACMG/ASHG position statement. I. Technical considerations.Genet Med 2000; 2: 356361.

30. Wapner RJ, Martin CL, Levy B, Ballif BC, Eng CM, ZacharyJM, Savage M, Platt LD, Saltzman D, Grobman WA, KlugmanS, Scholl T, Simpson JL, McCall K, Aggarwal VS, BunkeB, Nahum O, Patel A, Lamb AN, Thom EA, BeaudetAL, Ledbetter DH, Shaffer LG, Jackson L. Chromosomalmicroarray versus karyotyping for prenatal diagnosis. N EnglJ Med 2012; 367: 21752184.

31. Tabor A, Madsen M, Obel EB, Philip J, Bang J, Norgaard-Pedersen B. Randomised controlled trial of genetic amniocen-tesis in 4606 low-risk women. Lancet 1986; i: 12871292.

32. Alfirevic Z, Mujezinovic F, Sundberg K, Brigham S.Amniocentesis and chorionic villus sampling for prenataldiagnosis. Cochrane Database System Rev 2003; Issue 3:CD003252.

33. Dhallan R, Au WC, Mattagajasingh S, Emche S, Bayliss P,Damewood M, Cronin M, Chou V, Mohr M. Methods toincrease the percentage of free fetal DNA recovered from thematernal circulation. JAMA 2004; 291: 11141119.

34. Go AT, van Vugt JM, Oudejans CB. Non-invasive aneuploidydetection using free fetal DNA and RNA in maternal plasma:recent progress and future possibilities. Hum Reprod Update2011; 17: 372382.

35. Papageorgiou EA, Fiegler H, Rakyan V, Beck S, Hulten M,Lamnissou K, Carter NP, Patsalis PC. Sites of differential DNAmethylation between placenta and peripheral blood: molecularmarkers for noninvasive prenatal diagnosis of aneuploidies.Am J Pathol 2009; 174: 16091618.

36. Fan HC, Blumenfeld YJ, Chitkara U, Hudgins L, QuakeSR. Noninvasive diagnosis of fetal aneuploidy by shotgunsequencing DNA from maternal blood. Proc Natl Acad Sci US A 2008; 105: 1626616271.

37. Chiu RW, Chan KC, Gao Y, Lau VY, Zheng W, Leung TY, FooCH, Xie B, Tsui NB, Lun FM, Zee BC, Lau TK, Cantor CR,Lo YM. Noninvasive prenatal diagnosis of fetal chromosomalaneuploidy by massively parallel genomic sequencing of DNAin maternal plasma. Proc Natl Acad Sci U S A 2008; 105:2045820463.

38. Sehnert AJ, Rhees B, Comstock D, Comstock D, de FeoE, Heilek G, Burke J, Rava RP. Optimal detection offetal chromosomal abnormalities by massively parallel DNAsequencing of cell-free fetal DNA from maternal blood. ClinChem 2011; 57: 10421049.

39. Dan S, Wang W, Ren J, Li Y, Hu H, Xu Z, Lau TK, Xie J,Zhao W, Huang H, Xie J, Sun L, Zhang X, Wang W, Liao S,Qiang R, Cao J, Zhang Q, Zhou Y, Zhu H, Zhong M, GuoY, Lin L, Gao Z, Yao H, Zhang H, Zhao L, Jiang F, Chen F,Jiang H, Li S, Li Y, Wang J, Wang J, Duan T, Su Y, ZhangX. Clinical application of massively parallel sequencing-basedprenatal noninvasive fetal trisomy test for trisomies 21 and 18in 11,105 pregnancies with mixed risk factors. Prenat Diagn2012; 32: 12251232.

40. Fan HC, Quake SR. Sensitivity of noninvasive prenataldetection of fetal aneuploidy from maternal plasma usingshotgun sequencing is limited only by counting statistics. PLoSOne 2010; 5: e10439.

41. Liang D, Lv W, Wang H, Xu L, Liu J, Li H, Hu L, Peng Y, WuL. Non-invasive prenatal testing of fetal whole chromosomeaneuploidy by massively parallel sequencing. Prenat Diagn2013; 33: 409415.

42. Chiu RW, Sun H, Akolekar R, Clouser C, Lee C, McKernanK, Zhou D, Nicolaides KH, Lo YM. Maternal plasma DNAanalysis with massively parallel sequencing by ligation fornoninvasive prenatal diagnosis of trisomy 21. Clin Chem2010; 56: 459463.

43. Faas BH, de Ligt J, Janssen I, Eggink AJ, Wijnberger LD,van Vugt JM, Vissers L, Geurts van Kessel A. Non-invasiveprenatal diagnosis of fetal aneuploidies using massively parallelsequencing-by-ligation and evidence that cell-free fetal DNAin the maternal plasma originates from cytotrophoblastic cells.Expert Opin Biol Ther 2012; 12 (Suppl 1): S1926.

44. van den Oever JM, Balkassmi S, Verweij EJ, van Iterson M,Adama van Scheltema PN, Oepkes D, van Lith JM, HofferMJ, den Dunnen JT, Bakker E, Boon EM. Single moleculesequencing of free DNA from maternal plasma for noninvasivetrisomy 21 detection. Clin Chem 2012; 58: 699706.

45. Sparks AB, Wang ET, Struble CA, Barrett W, Stokowski R,McBride C, Zahn J, Lee K, Shen N, Doshi J, Sun M, Garrison J,Sandler J, Hollemon D, Pattee P, Tomita-Mitchell A, MitchellM, Stuelpnagel J, Song K, Oliphant A. Selective analysis of cell-free DNA in maternal blood for evaluation of fetal trisomy.Prenat Diagn 2012; 32: 39.

46. Dhallan R, Guo X, Emche S, Damewood M, Bayliss P, CroninM, Barry J, Betz J, Franz K, Gold K, Vallecillo B, Varney J.A non-invasive test for prenatal diagnosis based on fetal DNApresent in maternal blood: a preliminary study. Lancet 2007;369: 474481.

47. Ghanta S, Mitchell ME, Ames M, Hidestrand M, SimpsonP, Goetsch M, Thilly WG, Struble CA, Tomita-Mitchell A.Non-invasive prenatal detection of trisomy 21 using tandemsingle nucleotide polymorphisms. PLoS One 2010; 5: e13184.

48. Zimmermann B, Hill M, Gemelos G, Demko Z, Banjevic M,Baner J, Ryan A, Sigurjonsson S, Chopra N, Dodd M, LevyB, Rabinowitz M. Noninvasive prenatal aneuploidy testing ofchromosomes 13, 18, 21, X, and Y, using targeted sequencingof polymorphic loci. Prenat Diagn 2012; 32: 12331241.

49. Liao GJ, Chan KC, Jiang P, Sun H, Leung TY, Chiu RW,Lo YM. Noninvasive prenatal diagnosis of fetal trisomy21 by allelic ratio analysis using targeted massively parallelsequencing of maternal plasma DNA. PLoS One 2012; 7:e38154.

50. Poon LLM, Leung TN, Lau TK, Chow KC, Lo YMD.Differential DNA methylation between fetus and mother asa strategy for detecting fetal DNA in maternal plasma. ClinChem 2002; 48: 3541.

51. Tong YK, Ding C, Chiu RW, Gerovassili A, Chim SS, LeungTY, Leung TN, Lau TK, Nicolaides KH, Lo YM. Noninvasiveprenatal detection of fetal trisomy 18 by epigenetic allelicratio analysis in maternal pla

gine non-invasive prenatal testing for aneuploidy

Documents