Non-invasive prenatal testing for the prenatal screening of sex chromosome aneuploidies: A systematic review and meta-analysis of diagnostic test accuracy studies.

BACKGROUND: There is little evidence on the performance of non-invasive prenatal testing (NIPT) for the detection of fetal sex chromosomal imbalances. In this review, we aimed to appraise and synthesize the literature on the performance of NIPT for the prenatal detection of fetal sex chromosome aneuploidies.
METHODS: We performed our literature search in PubMed, Embase, Cochrane Library, Web of Science, and CADTH. Study selection and data extraction were performed by two reviewers independently. There were no restrictions on the study population. Meta-analyses were performed with "R" software. Pooled sensitivities and specificities with their 95% CI were estimated using a random-effects model. Heterogeneity between studies was assessed by a Q test.
RESULTS: Based on 11 studies in high prior risk pregnancies, including 116 affected fetuses in aggregate, Massively Parallel Shotgun Sequencing (MPSS) had a sensitivity of 93.9% (95% CI 84.1%, 97.8%) and a specificity of 99.6% (95% CI 98.7%, 99.9%) for the detection of 45,X. Based on four studies in high-risk pregnancies, with 83 affected fetuses in aggregate, Targeted Massively Parallel Sequencing (TMPS) had a sensitivity of 83.2% (95% CI 49.6%, 96.2%) and specificity was 99.8% (95% CI 98.3%, 100%) for the detection of 45,X. In mixed-risk pregnancies, the sensitivity of TMPS for the detection of 45,X was 90.9% (2 studies; 95% CI 70%, 97.7%) and specificity 99.9% (2 studies; 95% CI 99.4%, 100%); MPSS data were not available in such pregnancies. Based on smaller numbers of studies, and small numbers of affected fetuses in either high-risk or mixed-risk pregnancies (using either MPSS or TMPS), the sensitivities and specificities were equal to or greater than 76.2% for 47,XXX, 47,XXY and 47, XYY. The test failures for SCAs were 0.2% (95% CI 0%, 13.6%) for MPSS and 5.6% (95% CI 3.7%, 8.4%) for TMPS.
CONCLUSION: High-quality studies are still desirable in order to estimate the performance of NIPT for the detection of sex chromosome imbalances.

BACKGROUND

High‐income countries have prenatal screening programs for the detection of chromosomal anomalies of the fetus. However, the population target and the health conditions screened vary from one program to another (Benn et al., 2015).

Until recently, prenatal detection of chromosomal anomalies mainly relied on biochemical tests performed on pregnant women, ultrasounds of the fetus, and amniocentesis or chorionic villus sampling. Chromosomal anomalies can now be detected by a relatively new technology called non‐invasive prenatal testing (NIPT) that allows analyzing cell‐free placental DNA present in the maternal blood (Alldred et al., 2017; Bianchi & Chiu, 2018). In many health care jurisdictions, NIPT is generally offered as a screening test for the three most common autosomal trisomies (T21, T18, and T13), based on evidence of high sensitivity and specificity (Badeau et al., 2017).

NIPT could potentially be extended to offer the possibility of detecting other chromosomal conditions, such as sex chromosome aneuploidies (SCAs), that is, 45,X (Turner syndrome), 47,XXX (trisomy X syndrome), 47,XXY (Klinefelter syndrome), 47,XYY and 48,XXYY (Skuse et al., 2018). Some of these SCAs are associated with congenital anomalies. For instance, Turner Syndrome is associated with heart defects such as the coarctation of the aorta (Surerus et al., 2003), Klinefelter Syndrome could lead to hypogonadism, and cognitive impairment in males (Groth et al., 2013), and females with trisomy X syndrome have a high risk of suffering from infertility (Rafique et al., 2019). Knowing that a fetus is affected by SCAs can help parents and health care providers, therefore, to identify sooner what healthcare interventions are needed. This might benefit the health of the child. However, evaluations of NIPT for SCA differ by the type of technological platform used, nature of the population in which testing is performed, and the type of SCA included. Thus, the evidence is fragmented.

Objectives

The general objective of this work was to review the literature on the performance of NIPT for the prenatal detection of fetal sex chromosome aneuploidies, in order to identify the potential clinical value of including them in NIPT‐based prenatal screening programs.

METHODS

The study protocol was published in PROSPERO (registration number: CRD42019118785). We now report the study according to the PRISMA guideline (PRISMA, 2015).

Eligibility criteria

The eligibility criteria relied on the PICOS approach for diagnostic accuracy studies:

Participants/population: We took into consideration all pregnant women with singleton pregnancies whatever their risk level of having a child with an SCA or any of the three main autosomal abnormalities (T21, T18, T13) who provided a plasma sample for NIPT analysis and who also underwent to a confirmation test for fetal chromosomal anomalies.

Index test: The NIPT‐based molecular approach offered to the women was either targeted massively parallel sequencing (TPMS: SNP‐based) and/or massively parallel shotgun sequencing (MPSS: random throughout the genome).

Reference standard: The comparator was a karyotyping performed on samples obtained by amniocentesis, chorionic villus sampling or from the infant postnatally.

Target outcomes: The outcomes were sensitivity, specificity, and the test failure rate. The following reasons of test failure, fetal fraction DNA, sequencing process, quality control, assay failure were extracted.

Study design: Diagnostic test accuracy (DTA) studies were considered. We included non‐experimental cross‐sectional studies, and retrospective and prospective case‐control studies.

Studies for which it was not possible to extract or derive the number of true positives, false positives, false negatives, and true negatives were excluded.

Information sources

The literature search was performed in the following databases: PubMed, Embase, Cochrane Library, Web of Science, CADTH, and the references lists of published articles. The search was first performed on October 2, 2018, and updated on June 11, 2020. The literature search was not limited in time. We considered the literature from the first publication on NIPT for the detection of chromosomal imbalances. Only articles written in English were considered.

Search strategies

The relevant keywords that were used for the search are described in Table A1. Based on the PICOS approach, the search focused on five components: pregnancy, non‐invasive prenatal testing, test accuracy, reference test, and diagnostic test accuracy studies. An epidemiologist and a geneticist were consulted to improve the search strategy. The strategy was validated by a librarian. Moreover, the sensitivity of the search strategy was investigated by pre‐identifying some relevant studies and verifying if the search strategy was able to detect these studies.

TABLE A1

Search strategy

Database/consultation date	#	Concept	Search strategy	Results	Combination of concepts
					#	Total
PubMed/20200611	#1	Participant =Pregnancy	Pregnan*[TIAB] OR Gestation[TIAB] OR Fetus*[TIAB] OR “Fetal Structures” [TIAB] OR “Fetal Structure”[TIAB] OR “Fetal Tissue”[TIAB] OR “Fetal Tissues”[TIAB] OR “Pregnancy”[MeSH] OR “Pregnant Women”[MeSH] OR Fetus[MeSH]	1,110,096	#1 AND #2 AND #3 AND #4	2,034
	#2	Index test = Non‐invasive prenatal test	“Non‐invasive prenatal test”[TIAB] OR NIPT[TIAB] OR “Cell Free Nucleic Acid”[TIAB] OR “Circulating Nucleic Acids”[TIAB] OR “Cell Free DNA”[TIAB] OR cfDNA[TIAB] OR cirDNA[TIAB] OR “Cell Free Deoxyribonucleic Acid”[TIAB] OR “Circulating DNA”[TIAB] OR “Cell Free RNA”[TIAB] OR cfRNA[TIAB] OR cirRNA[TIAB] OR “Cell Free Ribonucleic Acid”[TIAB] OR “Circulating RNA”[TIAB] OR “Prenatal Diagnosis”[TIAB] OR “Prenatal Diagnoses”[TIAB] OR “Intrauterine Diagnosis”[TIAB] OR “Intrauterine Diagnoses”[TIAB] OR “Antenatal Diagnosis”[TIAB] OR “Antenatal Diagnoses”[TIAB] OR “Prenatal Screening”[TIAB] OR “Antenatal Screening”[TIAB] OR “Prenatal Diagnosis”[MeSH] OR “Cell‐Free Nucleic Acids”[MeSH]	88,361
	#3	Outcome 1 = Test accuracy	"test accuracy"[TIAB] OR sensitivity[TIAB] OR specificity[TIAB] OR “positive predictive value”[TIAB] OR “negative predictive value”[TIAB] OR “sensitivity and specificity”[MeSH] OR “Predictive Value of the Test”[TIAB]	1,436,981
	#4	Outcome 2 = Chromosome Disorders	“Chromosome Disorder”[TIAB] OR “Chromosome Disorders”[TIAB] OR “Chromosomal Disorder”[TIAB] OR “Chromosome Disorders”[TIAB] OR “Chromosome Aberration”[TIAB] OR “Chromosome Aberrations”[TIAB] OR “Chromosomal Aberration”[TIAB] OR “Chromosomal Aberrations”[TIAB] OR “Chromosome Abnormality”[TIAB] OR “Chromosome Abnormalities”[TIAB] OR “Chromosomal Abnormality”[TIAB] OR “Chromosomal Abnormalities”[TIAB] OR “Cytogenetic Abnormality”[TIAB] OR “Cytogenetic Abnormalities”[TIAB] OR “Cytogenetic Aberration”[TIAB] OR “Cytogenetic Aberrations”[TIAB] OR “Autosome Abnormality”[TIAB] OR “Autosome Abnormalities”[TIAB] OR Aneuploid*[TIAB] OR “chromosome imbalance”[TIAB] OR “chromosomal imbalance”[TIAB] OR "Chromosome Disorders"[MeSH] OR “Chromosome Aberrations”[MeSH] OR Aneuploidy[MeSH]	216,262
Embase/20200611	#1	Participant =Pregnancy	Pregnan*:ti,ab OR Gestation:ti,ab OR Fetus*:ti,ab OR “Fetal Structure*”:ti,ab OR “Fetal Tissue*”:ti,ab OR “pregnancy”/exp OR “pregnant woman”/exp OR Fetus/exp	1,224,399	#1 AND #2 AND #3 AND #4	1,919
	#2	Index test = Non‐invasive prenatal test	“Non‐invasive prenatal test*”:ti,ab OR NIPT:ti,ab OR “Cell Free Nucleic Acid*”:ti,ab OR “Circulating Nucleic Acid*”:ti,ab OR “Cell Free DNA”:ti,ab OR cfDNA:ti,ab OR cirDNA:ti,ab OR “Cell Free Deoxyribonucleic Acid”:ti,ab OR “Circulating DNA”:ti,ab OR “Cell Free RNA”:ti,ab OR cfRNA:ti,ab OR cirRNA:ti,ab OR “Cell Free Ribonucleic Acid”:ti,ab OR “Circulating RNA”:ti,ab OR “Prenatal Diagnos*”:ti,ab OR “Intrauterine Diagnos*”:ti,ab OR “Antenatal Diagnos*”:ti,ab OR “Prenatal Screening”:ti,ab OR “Antenatal Screening”:ti,ab OR “Prenatal Diagnosis”/exp OR “Cell‐Free Nucleic Acids”/exp	129,350
	#3	Outcome 1 = Test accuracy	“test accuracy”:ti,ab OR sensitivity:ti,ab OR specificity:ti,ab OR “positive predictive value*”:ti,ab OR “negative predictive value*”:ti,ab OR “predictive value of test*”:ti,ab OR “sensitivity and specificity”/exp OR “predictive value”/exp	1,551,179
	#4	Outcome 2 = Chromosome Disorders	“Chromosom* Disorder*”:ti,ab OR “Chromosom* Aberration*”:ti,ab OR “Chromosom* Abnormalit*”:ti,ab OR “Cytogenetic Abnormalit*”:ti,ab OR “Cytogenetic Aberration*”:ti,ab OR “Autosome Abnormalit*”:ti,ab OR Aneuploid*:ti,ab OR “chromosom* imbalance”:ti,ab OR “Chromosome Disorders”/exp OR “Chromosome Aberrations”/exp OR Aneuploidy/exp	255,660
Cochrane Library/20200611	#1	Participant =Pregnancy	(Pregnan*):ti,ab,kw OR (Gestation):ti,ab,kw OR (Fetus*):ti,ab,kw OR (“Fetal Structure*”):ti,ab,kw OR (“Fetal Tissue*”):ti,ab,kw OR [mh “pregnancy”] OR [mh “pregnant woman”] OR [mh “Fetus”]	66,902	#1 AND #2 AND #3 AND #4	23
	#3	Index test = Non‐invasive prenatal test	(“Non‐invasive prenatal test*”):ti,ab,kw OR (NIPT):ti,ab,kw OR (“Cell Free Nucleic Acid*”):ti,ab,kw OR (“Circulating Nucleic Acid*”):ti,ab,kw OR (“Cell Free DNA”):ti,ab,kw OR (cfDNA):ti,ab,kw OR (cirDNA):ti,ab,kw OR (“Cell Free Deoxyribonucleic Acid*”):ti,ab,kw OR (“Circulating DNA”):ti,ab,kw OR (“Cell Free RNA”):ti,ab,kw OR (cfRNA):ti,ab,kw OR (cirRNA):ti,ab,kw OR (“Cell Free Ribonucleic Acid*”):ti,ab,kw OR (“Circulating RNA”):ti,ab,kw OR (“Prenatal Diagnos*”):ti,ab,kw OR (“Intrauterine Diagnos*”):ti,ab,kw OR (“Antenatal Diagnos*”):ti,ab,kw OR (“Prenatal Screening”):ti,ab,kw OR (“Antenatal Screening”):ti,ab,kw OR [mh “Prenatal Diagnosis”] OR [mh “Cell‐Free Nucleic Acids”]	1,028
	#5	Outcome 1 = Test accuracy	(“test accuracy”):ti,ab,kw OR (sensitivity):ti,ab,kw OR (specificity):ti,ab,kw OR (“positive predictive value”):ti,ab,kw OR (“negative predictive value”):ti,ab,kw OR (“predictive value of tests”):ti,ab,kw OR [mh “sensitivity and specificity”] OR [mh “predictive value”]	65,443
	#7	Outcome 2 = Chromosome Disorders	(“Chromosom* Disorder*”):ti,ab,kw OR (“Chromosom* Aberration*”):ti,ab,kw OR (“Chromosom* Abnormalit*”):ti,ab,kw OR (“Cytogenetic Abnormalit*”):ti,ab,kw OR (“Cytogenetic Aberration*”):ti,ab,kw OR (“Autosome Abnormalit*”):ti,ab,kw OR (Aneuploid*):ti,ab,kw OR (“chromosom* imbalance”):ti,ab,kw OR [mh “Chromosome Disorders”] OR [mh “Chromosome Aberrations”] OR [mh “Aneuploidy”]	1,625
Web of Sciences/20200611	#1	Participant =Pregnancy	TS=(Pregnan* OR Gestation OR Fetus* OR “Fetal Structure*” OR “Fetal Tissue*”)	1,070,478	#1 AND #2 AND #3 AND #4	337
	#2	Index test =Non‐invasive prenatal test	TS=(“Non‐invasive prenatal test” OR NIPT OR “Cell Free Nucleic Acid*” OR “Circulating Nucleic Acid*” OR Cell “Free DNA OR cfDNA” OR cirDNA OR “Cell Free Deoxyribonucleic Acid*” OR “Circulating DNA OR Cell Free RNA” OR cfRNA OR cirRNA OR “Cell Free Ribonucleic Acid*” OR “Circulating RNA OR Prenatal Diagnos*” OR “Intrauterine Diagnos*” OR “Antenatal Diagnos*” OR “Prenatal Screening” OR “Antenatal Screening”)	9,049
	#3	Outcome 1 = Test accuracy	TS=(“test accuracy” OR sensitivity OR specificity OR “positive predictive value*” OR “negative predictive value*” OR “predictive value of tests”)	1,858,994
	#4	Outcome 2 = Chromosome Disorders	Ts=(“Chromosom* Disorder*” OR “Chromosom* Aberration*” OR “Chromosom* Abnormalit*” OR “Cytogenetic Abnormalit*” OR “Cytogenetic Aberration*” OR “Autosome Abnormalit*” OR Aneuploid* OR “chromosom* imbalance”)	107,251

Data management

The list of articles found was imported into EndNote (version 19.2.0.13018). Duplicates were identified and removed manually.

Selection of studies

A pilot selection was made by two reviewers (BS and CL). The pilot selection was carried out on 10% of the total number of articles identified as potentially eligible (320 of 3214 articles). Both researchers reviewed the titles and abstracts independently. Any article that got at least one "No" in the list of inclusion questions was excluded. The Kappa of the pilot selection between the two reviewers was ≥0.8, which was considered acceptable (McHugh, 2012). Thereafter, the two reviewers completed the selection of the studies independently with a kappa of 0.85. In case of a discrepancy, discussions were held in order to reach a consensus. A third reviewer (DR) was involved when a consensus could not be reached.

The full‐text study selection was performed by two reviewers (BS and CL) independently. Any disagreement between reviewers was settled, as for the title and abstract search, with the help of the third researcher (DR).

Data extraction

Our data extraction was performed on Review Manager 5.4. The same two reviewers as involved in title and abstract screening (BS and CL) extracted the data from all included studies independently. Our data extraction included the following items:

Reference details enabling the publication to be accessed (first author, publication year, etc.);

Study design: retrospective/prospective, blinded/unblinded, case‐control/non‐case‐control;

Population characteristics: total number of pregnant women, prior risk.

Features of the index test:

Test‐specific details: TMPS or MPSS, cutpoint, and test failure rate.

Features of the reference standard: fetal karyotyping, diagnostic sample.

Outcome characteristics:

Information on specific types of SCAs: the total number of cases of SCAs;

Test performance: true positive, false positive, true negative, false negative, and test failure.

All extracted data were cross‐checked. Disagreements were solved by discussion and search of a consensus between the two reviewers (BS and CL). A third reviewer (DR) was involved in the case a consensus could not be reached.

Risk of bias and applicability

The QUADAS‐2 tool for the assessment of diagnostic accuracy study was used to assess the quality of the methodology of the included studies (Whiting et al., 2011). This was the basis for a series of signaling questions (see Table A2) each of which was answered as "yes," "no," or "unclear." The reason for each judgment on the risk of bias and applicability was recorded. A study was classified as "low risk of bias" when a "yes" to all signaling questions was recorded. It was classified as of "high risk of bias" when a "no" was recorded for one or more signaling questions. It was classified as "unclear risk of bias" when an "unclear" for one or more signaling questions were recorded. In addition, the study was classified as “ high risk of bias” when at least a “no” and an “unclear” were recorded. Applicability concerns were judged according to our review questions. Each study was classified as "low, high, or unclear concern." A study that had insufficient information was classified as "unclear concern." The two reviewers (BS and CL) applied the QUADAS‐2 tool to each included study independently. A pilot assessment was performed with three studies. Disagreements were resolved by discussion and search for a consensus. A third reviewer (FR) was consulted in case of doubt.

TABLE A2

QUADAS‐2 tool for assessing the methodological quality of included studies

Domain	Patients selection	Index test	Reference standard	Flow and timing
Signaling Question (yes, no, or unclear)	Was a consecutive or random sample of patients enrolled?	Were the index test results interpreted without knowledge of the results of the reference standard?	Is the reference standard likely to correctly classify the target condition?	Was there an appropriate interval between gNIPT and reference standard?
	Was a case‐control design avoided?	If a threshold was used, was it pre‐specified?	Were the reference standard results interpreted without knowledge of the results of the index test?	Did all analyzed patients receive the reference standard?
	Did the study avoid inappropriate exclusions?			Were all patients included in the analysis?
Risk of bias (high, low or, unclear)	Could the selection of patients have introduced bias?	Could the conduct or interpretation of the index test have introduced bias?	Could the reference standard, its conduct, or its interpretation have introduced bias?	Could the patient flow have introduced bias?
Concerns about Applicability (high, low or, unclear)	Are there concerns that the included patients and setting do not match the review question?	Are there concerns that the index test, its conduct, or interpretation differ from the review question?	Are there concerns that the target condition as defined by the reference standard does not match the review question?

Statistical analyses and data synthesis

Review Manager 5.4 was used to synthesize the qualitative evidence of the included studies. Each included study was systematically and comprehensively assessed. The features of the studies, including study population, index test, reference standard, and outcome, were described.

The main statistical analyses aimed to examine the performance of the NIPT to detect target SCAs, both individually and globally for each condition. Two‐by‐two tables were constructed with the main outcomes (true positive, false positive, true negative, and false negative) in order to calculate the individual studies’ sensitivity, specificity, and false‐positive rate. True positives consisted of the fetus with the SCAs under analysis. True negatives consisted of unaffected fetus and fetus with chromosomal abnormalities other than the ones under analysis.

Meta‐analyses were performed with R software (version 3.6.1) when there were two or more studies on a specific chromosomal anomaly detected using a specific platform. When there were three or more studies, we performed a leave‐out‐one‐study sensitivity analysis in order to determine how each individual study might affect the overall estimates. The pooled sensitivities and specificities with their 95% CI were estimated using a random‐effects model because we assumed that all included studies in our meta‐analyses were a random sample of all possible studies that might meet our inclusion criteria and because the performance of NIPT might vary across studies due to differences in study populations and study protocols (Riley et al., 2011). Hierarchical summary receiver operation (HSROC) model was performed in some cases in order to take into account the correlation between sensitivity and specificity. The heterogeneity among the studies was assessed by the Q test and was quantified by an I‐square test. Furthermore, the between‐studies variation was estimated by a tau‐square test.

We performed two sub‐group analyses regarding the prior risk (high vs. mixed) of having a child with aneuploidy of any type and NIPT‐based methods (MPSS vs. TMPS) separately because it was expected that the performance of NIPT would be different.

RESULTS

Search

The PRISMA Flow Diagram for the selection of studies is presented in Figure 1. After the elimination of duplicates, 3214 articles were identified, of which 185 met the eligibility criteria based on the titles and the abstract. A review of the full text of these 185 articles resulted in the inclusion of 21 articles; reasons for exclusion of the other 164 articles are summarized in Figure 1.

FIGURE 1

PRISMA Flow Diagram for the selection of studies

Basic features of included articles

Studies were classified into two groups: massively parallel shotgun sequencing (MPSS) (15 studies) and targeted massively parallel sequencing (TMPS) (6 studies). The studies were conducted both retrospectively and prospectively. The majority of the studies used Illumina sequencing to screen for chromosome abnormalities. All of the studies used fetal karyotype as the reference for SCAs (Table 1). Details regarding the performance of individual studies for each condition in high‐risk pregnancies are summarized in Figures 2 and 3 and in mixed‐risk pregnancies in Figure 4. In addition, studies were classified into two groups; high and mixed risk pregnancy.

TABLE 1

Basic features of included studies, N = 21

Study ID	Condition a	Study design and participants	Prior risk	Index test	Threshold	Reference standard
Massively Parallel Shotgun Sequencing (MPSS), N = 15
Bianchi (2012)	45,X	Prospective, multicenter observational study 532 singleton pregnancies were analyzed	High	Illumina HiSeq 2000 sequencer in 6‐plex	NCV chrom X < −4.0 and NCV chrom Y < 2.5	Karyotype
Bianchi (2013)	45,X	Retrospective study 113 singleton pregnancies were analyzed	High	Illumina TrueSeq 3.0 sequencing chemistry	NCV chrom X < −3.0 and NCV chrom Y < 3.0	Karyotype
Ivashchenko (2019)	47,XXX	Observational study 149 singleton pregnancies were analyzed	High	Next‐Generation Sequencing	Z‐score <−3 or >3	Karyotype, FISH
Jiang (2012)	45,X, 47,XXY, 47,XYY	Prospective Case‐Control design 903 singleton pregnancies were analyzed	High	Illumina Genome Analyzer IIx or HiSeq 2000 sequencer in multiplex	t score <−2.5 for 47,XXX and 45,X t score >2.5 for 47,XXY and 47,XYY	Karyotype
Lau (2012)	45,X, 47,XXX, 47XXY, 47,XYY	Blinded prospective cohort study 108 singleton pregnancies were analyzed	High	Illumina HiSeq 2000 sequencer in 12‐plex	Z‐score of [−3.5; −2.5] for 45,X, and 47,XXX ‐Z‐score of [2.5;3.5] for 47,XXY, and 47,XYY	Karyotype
Liu (2012)	45,X, 47,XXX, 47XXY, 47,XYY	Prospective cohort study 153 singleton pregnancies were analyzed	High	Illumina HiSeq sequencer in multiplex	Z score ≥3	Karyotype
Liang (2013)	45,X, 47,XXX, 47,XXY, 47,XYY	Blinded prospective study 412 singleton pregnancies were analyzed	High	v2 Illumina HiSeq 2000 sequencer in 8‐plex or 12‐ plex	Z‐score of −2.91 for 45,X, and 47,XXX Z‐score of 3 for 47,XXY, and 47,XYY	Karyotype
Lefkowitz (2016)	45,X, 47,XXX, 47,XXY, 47,XYY	Blinded retrospective clinical study 1,166 singleton pregnancies were analyzed	High	Illumina HiSeq 2000 sequencer in 6‐plex or uniplex	Z‐score of [−3.5; −2.5] for 45,X, and 47,XXX Z‐score of [2.5;3.5] for 47,XXY, and 47,XYY	G‐band karyotype
Li (2018)	45,X, 47,XXX, 47,XXY, 47,XYY	Retrospective observational study 570 singleton pregnancies were analyzed	High	BGISEQ‐500 sequencer	Z‐score <−3 or >3	Karyotype
Mazloom (2013)	45,X, 47,XXX, 47,XXY, 47,XYY	Blinded prospective cohort study 411 singleton pregnancies were analyzed	High	Illumina v3 HiSeq 2000 sequencer in 12‐plex	‐Z‐score of [−3.5; −2.5] for 45,X, and 47,XXX ‐Z‐score of [2.5;3.5] for 47,XXY, and 47,XYY	Karyotype
Porreco (2014)	45,X, 47,XXX, 47,XXY, 47,XYY	Blinded prospective multicenter observational study 3,278 singleton pregnancies were analyzed for 45,X, 47,XXX, and 3,201 were analyzed for 47,XXY, 47,XYY	High	Illumina HiSeq 2000 sequencer in 12‐plex	‐Z‐score of [−3.5; −2.5] for 45,X, and 47,XXX ‐Z‐score of [2.5;3.5] for 47,XXY, and 47,XYY	Karyotype
Sehnert (2011)	45,X	Retrospective cohort study 47 singleton pregnancies were analyzed	High	Illumina Genome Analyzer IIx sequencer in uniplex	NCV <−3	Karyotype
Shaw (2014)	45,X, 47,XXX, 47,XXY, 47,XYY	Prospective cohort study 200 singleton pregnancies were analysed	Mixed	v2 Illumina HiSeq 2000 sequencer in 12‐plex	Z‐score of [−3.5; −2.5] for 45,X, and 47,XXX ‐Z‐score of [2.5;3.5] for 47,XXY, and 47,XYY	Karyotype
Song (2015)	45,X, 47,XXX, 47,XXY	Blinded prospective cohort study 178 singleton pregnancies were analyzed	High	Illumina v2 HiSeq 2000 sequencer in 12‐plex	Z‐score of −3 for 45,X and 47,XXX ‐Z‐score of 3 for 47,XYY	Karyotype
Zhu (2019)	45,X, 47,XXY	Blinded retrospective observational study 802 singleton pregnancies were analyzed	High	Next‐Generation Sequencing (Ion Proton sequencing)	Z absolute <1.96 or >3	Karyotype
Targeted Massively Parallel Sequencing (TMPS), N = 6
Hooks (2014)	45,X, 47,XXX, 47,XXY, 47,XYY	Prospective study 414 singleton pregnancies were analyzed	High	DANSR™ Assay and algorithm (FORTETM)	NR	Karyotype
Nicolaides (2013)	45,X, 47,XXX, 47,XXY, 47,XYY	Blinded prospective study 229 singleton pregnancies were analyzed	High	SNP‐based Method (NATUS algorithm), Illumina Genome Analyzer IIx or HiSeq sequencer, 19,488‐plex targeted PCR	NR	Karyotype
Nicolaides (2014)	45,X, 47,XXX, 47,XXY, 47,XYY	Case‐control study 172 singleton pregnancies were analysed	High	DANSR™ assay (FORTE), Illumina HiSeq 200 in 96‐plex	FORTE risk score of 1%	Karyotype
Pergament (2014)	45,X	A blinded prospective cohort study 963 singleton pregnancies were analyzed	Mixed	The Next‐generation Aneuploidy Test Using SNPs algorithm	NR	Karyotype
Persico (2016)	45,X, 47,XXX, 47,XXY, 47,XYY	A blinded prospective cohort study 249 singleton pregnancies were analyzed	High	SNP‐based Method (NATUS algorithm), Illumina Genome Analyzer IIx or HiSeq sequencer, 19,488‐plex targeted PCR	A risk score of 1%	Karyotype
Samango‐Sprouse (2013)	45,X, 47,XXX, 47,XXY, 47,XYY	A blinded prospective cohort study 286 singleton pregnancies were analyzed	Mixed	SNP‐based Method (NATUS algorithm), Illumina HiSeq sequencer, 19,488‐plex targeted PCR	NR	Karyotype

Abbreviation: NCV, normalized chromosome value.

The conditions presented in this review.

FIGURE 2

Performance of MPSS in the detection of 45,X, 47,XXX, 47,XXY, and 47,XYY in high‐risk pregnancy

FIGURE 3

Performance of TMPS in the detection of 45,X, 47,XXX, 47,XXY, and 47,XYY in high‐risk pregnancy

FIGURE 4

Performance of MPSS and TMPS in the detection of 45,X, 47,XXX, 47,XXY, and 47,XYY in mixed‐risk pregnancy

Performance of NIPT in high‐risk pregnancy

45,X syndrome

MPSS platform

A total of 11 studies were included in the meta‐analysis on the MPSS platform. The sample consisted of 116 affected and 8224 unaffected fetuses. Sensitivities in individual studies ranged from 63% to 100% (Figure 2). The pooled sensitivity was 93.9% (95% CI 84.1%–97.8%). The pooled specificity was 99.6% (95% CI 98.7%–99.8%), with high heterogeneity (Table 2).

TABLE 2

HSROC model for high‐risk pregnancy

Performance	45,X syndrome (MPSS) (Affected = 116, Unaffected = 8224)	47,XXX (TMPS) (Affected = 6, Unaffected = 580)	47,XXY (TMPS) (Affected = 8, Unaffected = 827)
Sensitivity (95% CI)	93.9% (84.1%, 97.8%)	76.2% (36%, 94.8%)	82.9% (38.2%, 97.4%)
Specificity (95% CI)	99.6% (98.7%, 99.9%)	99.5% (98.4%, 99.8%)	99.9% (99.1%, 100%)

TMPS platform

The meta‐analysis was performed on four studies. The sample consisted of 83 affected and 985 unaffected fetuses. The pooled sensitivity and specificity were 83.2% (95% CI 49.6%–96.2%), with considerable heterogeneity and 99.8% (95% CI 98.3%–100%), respectively (Figure 5, 6).

FIGURE 5

Forest plot of the sensitivity of TMPS for 45,X in high‐risk pregnancy

FIGURE 6

Forest plot of the specificity of TMPS for 45,X in high‐risk pregnancy

47,XXX syndrome (triple X syndrome)

A total of five studies were included in a meta‐analysis on the MPSS platform. The sample consisted of 16 affected and 5302 unaffected fetuses. The pooled sensitivity was 99.5% (95% CI 0.1%–100%), with high heterogeneity (Figure 7). The pooled specificity was 99.9% (95% CI 99.8%–100%) (Figure 8).

FIGURE 7

Forest plot of the sensitivity of MPSS for 47,XXX in high‐risk pregnancy

FIGURE 8

Forest plot of the specificity of MPSS for 47,XXX in high‐risk pregnancy

Two studies were analyzed. The sample consisted of 6 affected and 580 unaffected fetuses. The pooled sensitivity was 76.2% (95% CI 36%–94.8%). The pooled specificity was 99.5% (95% CI 98.4%–99.8%) (Table 2).

47,XXY syndrome (Klinefelter syndrome)

Eight studies were included in the meta‐analysis. The sample consisted of 24 affected and 7597 unaffected fetuses. The pooled sensitivity was 100% (95% CI 0%–100%), with high heterogeneity (Figure 9). The pooled specificity was 100% (95% CI 99.6%–100%) (Figure 10).

FIGURE 9

Forest plot of the sensitivity of MPSS for 47,XXY in high‐risk pregnancy

FIGURE 10

Forest plot of the specificity of MPSS for 47,XXY in high‐risk pregnancy

A total of three studies were analyzed. The sample consisted of 8 affected and 827 unaffected fetuses. The pooled sensitivity and specificity were 82.9% (95% CI 38.2%–97.4%) and 99.9% (95% CI 99.1%–100%), respectively (Table 2).

47,XYY syndrome

A total of six studies were analyzed. The sample consisted of 12 affected and 6373 unaffected fetuses. The pooled sensitivity was 91.7% (95% CI 58.7%–98.8%) (Figure 11). The pooled specificity was 100% (95% CI 99.9%–100%) (Figure 12).

FIGURE 11

Forest plot of the sensitivity of MPSS for 47,XYY in high‐risk pregnancy

FIGURE 12

Forest plot of the specificity of MPSS for 47,XYY in high‐risk pregnancy

Based on a single study, consisting of 3 affected and 169 unaffected fetuses. The sensitivity for the detection of 47,XYY syndrome was 100% (95% CI 29% to 100%). The specificity was 100% (95% CI 99.8%–100%) (Figure 3).

Performance of NIPT in mixed‐risk pregnancy

Based on two studies, consisting of 12 affected and 296 unaffected fetuses. The sensitivity and specificity ranged from 75% to 100% (95% CI 19%–100%) and 100% (95% CI 96%–100%), respectively (Figure 4).

Two studies were analyzed. The sample consisted of 22 affected and 1128 unaffected fetuses. The pooled sensitivity and specificity was 90.9% (95% CI 70%–97.7%) (Figure 13) and 99.9% (95% CI 99.4%–100%), respectively (Figure 14).

FIGURE 13

Forest plot of the sensitivity of TMPS for 45,X in mixed‐risk pregnancy

FIGURE 14

Forest plot of the specificity of TMPS for 45,X in mixed‐risk pregnancy

47,XXY syndrome (Klinefelter syndrom)

According to the result of one study, performing with one affected and 107 unaffected fetuses, the sensitivity of the MPSS platform for the detection of 47,XXY syndrome was 100% (95% CI 3%–100%) and its specificity was 100% (95% CI 97%–100%) (Figure 4).

Based on one study, conducted with two affected and 184 unaffected fetuses, the sensitivity was 100% (95% CI 16%–100%), and the specificity was 100% (95% CI 98%–100%) (Figure 4).

The sensitivity and specificity of the MPSS platform for the detection of 47,XYY syndromes were 100% (95% CI 3% to 100%) and 100% (95% CI 98%–100%), respectively, according to the result of a single study, conducting with one affected and 199 unaffected fetuses (Figure 4).

The result of one study conducting with 1 affected and 185 unaffected fetuses showed that the sensitivity of the TMPS platform was 100% (95% CI 3%–100%). The specificity was 100% (95% CI 98%–100%) (Figure 4).

Test failure

In these analyses, we included only studies that reported the test failure rate. Many studies were, therefore, excluded.

MPSS platform

Based on eight and three studies, the test failure rate was 0.3% (95% CI 0%–1.6%) for all aneuploidies (Figure 15) and 0.2% (95% CI 0%–13.6%) for SCAs (Figure 16), respectively.

FIGURE 15

Forest plot of the test failure rate of MPSS for all aneuploidy.

FIGURE 16

Forest plot of the test failure rate of MPSS for sex chromosome aneuploidy

TMPS platform

Two studies were included; the test failure rate was 4.6% (95% CI 3.1%–6.8%) for all aneuploidies (Figure 17). Besides, based on four studies, the test failure rate for SCAs was 5.6% (95% CI 3.7%–8.4%) (Figure 18).

FIGURE 17

Forest plot of the test failure rate of TMPS for all aneuploidy

FIGURE 18

Forest plot of the test failure rate of TMPS for sex chromosome aneuploidy

Sensitivity analyses

The results of the sensitivity analyses for TMPS and MPSS platforms are presented in Tables 3 and 4, respectively. Using the leave‐out‐one study approach resulted in little change in the pooled estimates of clinical sensitivities and specificities.

TABLE 3

Sensitivity analysis of MPSS platform in high‐risk pregnancy

Performance	45,X syndrome	Sensitivity analysis	47,XXX syndrome	Sensitivity analysis	47,XXY syndrome	Sensitivity analysis	47,XYY syndrome	Sensitivity analysis
Sensitivity (95% CI)	93.9% (84.1%, 97.8%)	93.6% (82.3%, 98%)	99.5% (0.1%, 100%)	100% (0%, 100%)	100% (0%, 100%)	100% (0%, 100%)	91.7% (58.7%, 98.8%)	92.1% (0%, 100%)
Specificity (95% CI)	99.6% (98.7%, 99.9%)	99.2% (97.3%, 99.8%)	99.9% (99.8%, 100%)	99.2% (99.8%, 100%)	100% (99.8%, 100%)	100% (98.4%, 100%)	100% (99.9%, 100%)	100% (99.8%, 100%)

TABLE 4

Sensitivity analysis of TMPS platform in high‐risk pregnancy

Performance	45,X syndrome	Sensitivity analysis	47,XXY syndrome	Sensitivity analysis
Sensitivity (95% CI)	83.2% (49.6%, 96.2%)	86.5% (45%, 97.7%)	82.9% (38.2%, 97.4%)	76.7% (9.4%, 99.3%)
Specificity (95% CI)	99.8% (98.3%, 100%)	99.8% (0%, 100%)	99.9% (99.1%, 100%)	99.9% (0%, 100%)

Quality assessment of the studies

The methodology of each included study was assessed for its potential biases and applicability concerns. The results of this assessment are presented in Figures 19 and 20. Based on QUADAS‐2 criteria, all studies were classified as high risk of bias. All 21 studies (100%) were classified as high risk of bias regarding their patient selection. Three studies (14.3%) were considered to have a high applicability concern.

FIGURE 19

Risk of bias and applicability concerns of the studies based on the QUADAS‐2 criteria

FIGURE 20

Summary of risk of bias and applicability concerns of the studies based on the QUADAS‐2 criteria

DISCUSSION

Principal findings

Twenty‐one studies were included in this review; 15 studies were conducted with the MPSS platform and six studies with the TMPS platform. Massively Parallel Shotgun Sequencing (MPSS) had a sensitivity of 93.9%, 99.5%, 100% and 91.7% and had a specificity of 99.6%, 99.9%, 100% and 100%, respectively, for the detection of 45,X, 47,XXX, 47,XXY and 47,XYY in high prior risk pregnancies. Targeted Massively Parallel Sequencing (TMPS) had a sensitivity of 83.2%, 76.2%, 82.9% and 100%. Its specificity was 99.8%, 99.5%, 99.9%, and 100%, respectively, for the detection of these SCAs in the high‐risk pregnant women. The test failure of NIPT on all aneuploidy was 0.3% and 4.6% for MPSS and TMPS, respectively. Meanwhile, the test failure rates for SCAs exclusively were 0.2% and 5.6%, respectively.

Interpretation of the findings

The performance of NIPT for the detection of common aneuploidies (trisomy 21, 18, and 13) has been well studied, and NIPT is now considered as accurate and cost‐effective for these chromosomal conditions when used as a 2nd‐tier screening test (Badeau et al., 2017; Gil et al., 2017). The test is non‐invasive, more sensitive, and specific than traditional screening tests performed with biochemical markers and ultrasounds. One wonders now if it should not be used for the screening of conditions beyond T21, 13, and 18 (Alldred et al., 2012).

The results of this review confirm and extend the findings of another review performed on NIPT for the detection of common sex chromosome aneuploidies (Badeau et al., 2017), especially the NIPT failure rate for SCAs. The accuracy of the test depended on the conditions searched and the NIPT method. According to our meta‐analysis, both platforms have an acceptable sensitivity for the detection of 45,X in high‐risk pregnancy. However, these results are based on studies that might contain biases. For the other SCAs, that is, 47,XXX, 47,XXY, and 47,XYY, more quality studies are still needed. From a clinical perspective, effectively, having a clear idea of the performance of NIPT, hence its added‐value, is of interest. Knowing that a fetus is affected with an SCA can help parents and health care providers to better plan relevant healthcare interventions, for instance, fetal echocardiograms to make sure that there are no cardiac abnormalities in fetuses with 45,X (Turner syndrome) (Surerus et al., 2003) or blood tests to diagnose hypogonadism in children with a 47,XXY condition (Klinefelter syndrome). The detection at a prenatal stage allows the implementation of interventions such as testosterone supplements earlier on after birth, which can improve the health outcome of affected males (Groth et al., 2013).

Yet, in many health care jurisdictions, an economic evaluation is an important step and has to be performed before considering expanding the conditions listed in a NIPT‐based screening program. A current study showed that adding SCAs in a first‐tier NIPT‐based program increased the screening cost (Xie et al., 2020). However, due to the low external validity of the economic evaluation, economic evaluations should be performed in every health care jurisdiction.

Our study aimed to compare the performance of MPSS versus TMPS One significant difference was the test failure rate of NIPT additional to SCAs which was lower in MPSS than TMPS, 0.2% (95% CI 0% to 13.6%) versus 5.6% (95% CI 3.7%, 8.4%). One major factor in explaining this difference was that MPSS included studies that did not measure fetal fraction (Palomaki & Kloza, 2018), while the main cause of test failure is known to be a low fetal fraction (Pergament et al., 2014). Recently, a review published their results in terms of failure rate (ranged from 0.9% to 5.6%) ("Noninvasive Prenatal Testing for Trisomies 21, 18, and 13, Sex Chromosome Aneuploidies, and Microdeletions: A Health Technology Assessment," 2019), nevertheless, they did not do a subgroup analysis (MPSS vs. TMPS). Therefore, we cannot compare our results directly with them. In addition, one study reported a 0.2% additional failure rate due to 45,X (Pergament et al., 2014).

In brief, our results of suggesting that clinicians should be cautious about not overstating the benefits of NIPT for the detection of SCAs.

Strengths of the study

The first strength lies in the fact that the protocol had been validated by experts and published in PROSPERO (March 19, 2019). The search for data was performed in relevant databases without any restriction on time, NIPT methods, SCAs, and population. Selection of studies, data extraction, and quality assessment of studies’ methodology was performed independently by two reviewers.

The second strength of the study is the strict use of karyotyping as the gold standard for the diagnosis of an SCA.

The last strength is related to the data analyses. We performed a stratified‐analysis on two NIPT platforms, and different risk groups of pregnancy, allowing us to explore and compare their performance for the detection of SCAs. Besides, we strengthened the results of our meta‐analysis by performing a sensitivity analysis using a leave‐one‐out approach. This analysis suggested that our results were not due to a single study.

Limitations of the study

The first limitation of this study comes from the fact that there was a limited number of studies for 47,XXX, 47,XXY and 47,XYY, especially when the TMPS platform was used. Our meta‐analyses performed for the above conditions also were also limited by the number of affected cases. Consequently, the 95% CI of the sensitivities of NIPT to detect those conditions were wide.

The second limitation is related to the quality of the studies’ methodology. The QUADAS‐2 criteria suggested that all of the studies included in this review were at high risk of bias. Therefore, our review might fall under or overestimated the performance of NIPT for the screening of SCAs.

The third limitation is the heterogeneity of the included studies. In Table 1, we can see some differences between studies, for instance, regarding the cutoff points and index tests. There was a high I‐square score at the forest plots. The high I‐square score suggests the presence of heterogeneity between the included studies. Nevertheless, when the number of studies is small, the I‐square test might suggest heterogeneity when there is no variation in the outcome considered between the studies or a homogeneity when there is large heterogeneity between studies (von Hippel, 2015).

CONCLUSIONS

Implication for practice

Based on the results of our meta‐analysis MPSS based‐NIPT can be considered as performant for the detection of 45,X in high‐risk pregnancies. However, this result is based on studies that might contain biases. Moreover, for the other SCAs including 47,XXX, 47,XXY, and 47,XYY, more quality studies are still needed to include in the meta‐analysis in order to ensure the statistical power of the meta‐analysis.

Implication for research

The results of our review showed that the MPSS NIPT‐based platform had a better performance to detect SCAs in high‐risk pregnancies. Moreover, its test failure rate for SCAs was lower compared to TMPS. Nevertheless, one should note that all of the studies included in the meta‐analysis were susceptible to bias. Therefore, future research should ensure to avoid biases that might occur.

CONFLICT OF INTEREST

The authors confirm that there was no conflict of interest in this project.

AUTHORS’ CONTRIBUTIONS

DR initiated the project. BS drafted the protocol. The protocol was approved by all the authors. BH and CL performed the study's selection, data extraction and assessed the methodological quality of the included studies. SL and FR provided their expertise to validate the clinical and laboratory information to be extracted. JL provided his expertise to validate the statistical approach. BH analyzed the data and finalized the review. All authors critically commented on the manuscript. The manuscript was approved by all authors.