Results

Search results
A total of 1209 articles were retrieved from the initial database searches, and 512 duplicates were eliminated. Following the screening of titles/abstracts, 624 publications were excluded and 61 articles were retrieved for full review and eligibility assessment. Finally, 14 articles were considered eligible for inclusion (see the PRISMA flowchart; Figure S1).
Included study characteristics
Table S1 summarizes the methodological characteristics of the included studies. Of the 14 studies included in this review, 13 were retrospective and 1 was prospective. Of the 13 retrospective studies, 12 consisted of observational or retrospective cohort studies, and 1 was a nested case-control study The cohort size of the studies ranged from 200 to 93 048. The maternal age was unknown in 2 studies and the interquartile range was the only information given in one study. Ten studies used massive parallel sequencing (MPS) technology as the DNA sequencing method. All the positive NIPT screening cases were supported by prenatal or postnatal karyotyping, provided that the participants consented to the procedure, and they were not lost to follow-up. In 4 studies, both autosomal trisomies and SCAs were studied, whereas only autosomal trisomies were studied in 4 articles, and only SCAs in 6 articles. All in all, 8 studies for autosomal trisomies and 10 for SCAs were gathered.
Methodological quality of included studies
The CASP tool was used to assess the risk of bias. A summary of the results is presented in Table S2 and described in Appendix S2 .
Analysis of Autosomal Trisomies: 21, 18, and 13
Out of the 14 included studies, 8 studies evaluated the autosomal trisomies 21, 18 and 13. The results are summarised in Table 1 .
The cohort sizes range between 200 and 40311 participants (Table S1 ). Of the 8 studies, the highest prevalence of a positive cffDNA test result found for T21 was of 5.5% (Table 3).
In general the studies concurred in that T21 was found to have the highest PPVs as compared to T18 and T13 (Table 1). Three studies showed a 100% PPV for T21. Another 2 studies reported PPVs between 90-99% whilst PPVs of 80% to 90% were seen in 2 studies.
The lowest relative overall PPVs were reported for T13. A PPV of 0%, followed by 14.3% were the lowest value reported across the 8 studies. Results were quite heterogenous, with PPVs as high as 100%, and a number of results between 20-90% also reported17,19,22. One study reported no cases of trisomy 13, and consequently did not present its PPV.
For T18, the results were intermediate between those of T21 and T13. Lee DE et al. (2019) and Serapinas D et al. (2020) reported PPVs of 100%. Alyafee Y et al. (2021) found an 80% PPV. The remaining 4 studies had results below 60%: 59%, 58.7%, 48.2% and 0%.
Among the 8 studies that reported on the autosomal trisomies, 1 study omitted the individual PPVs and only included the overall PPV of the 3 trisomies, with a result of 64.7%16. However, a PPV of 100% was reported when the cases with multiple clinical indications, rather than just one, and including abnormal ultrasound findings for NIPT, were exclusively accounted for.
Analysis of Sex Chromosome Aneuploidies
Out of the 14 included studies, 10 studies involved analysis of SCAs. The results are summarised in Table 2 .
The number of participants per study ranged from 862 to 93048 (Table S1 ). The prevalence of a high-risk NIPT is presented individually for each SCA in 6 studies, while only the overall PPV for all SCAs is available for 4 studies (Table 4). Of all these studies, the highest prevalence of a positive cffDNA test is 0.6%.
In comparison to the other SCAs, the lowest PPVs were associated with Turner syndrome (all PPVs were <30%). The PPVs for Klinefelter syndrome ranged from 25% - 77.8%. In the case of triple X syndrome, 4 studies reported PPVs from 29.7% - 51.7%, whilst two studies reported a PPV of 100%. Finally, studies which screened for Jacob’s syndrome demonstrated heterogenous results. Two studies returned PPVs of 100%, with others in the range of 60% - 83.3%. One study returned a low PPV of 27.3%. It is pertinent to note that of the two studies reporting a PPV of 100%, the latter study only identified one case that bore the 47,XYY SCA.
Among the 10 studies that reported on SCAs, 3 studies reported only the overall PPV, without specifying the PPV of each individual SCA. They all presented PPVs < 50%.
Analysis of Discordant (false positives; FP and false negative; FN) results
The incidence of FP and FN cases were not uniformly reported across the 14 studies.
Although the number of FP cases was documented in 13 of the studies, the cause of the result was only confirmed in 4 studies, while the others offered a generic (non-specific) explanation. Ma L et al. (2018) proved that a case wrongly screened as XYY was later confirmed as a microdeletion. Placental mosaicism was a cause for one T18 FP case, whilst one T13 FP case was confirmed postnatally by placental investigation and karyotyping of cord blood, which explains the 0% PPVs found in this study for T18 and T13. Furthermore, 2 maternal mosaicism cases were confirmed by maternal karyotyping in a 2021 study. Two 47,XXX FP cases were identified, secondary to undiagnosed maternal 47,XXX SCA, as confirmed by maternal blood karyotype. The final study presented PPVs of 100% across all the identified aneuploidies, and therefore no FP were detected.
In general, FN were poorly reported, or not reported at all in most of the studies. A small number of studies proffered some basic explanation (without differentiating between FP and FN), whilst others specifically described the FN outcomes of their studies. Our analysis shows that no FN were identified in 5 studies. However, placental mosaicism was confirmed as a source of FN in a single study, in which 3 cases of T18 were identified by CVS with quantitative fluorescent polymerase chain reaction.
Inconclusive results
Reporting of the percentage of no-call (or inconclusive) results was absent in 5 studies. Nevertheless, 10 of the 14 studies stated that the main cause for no-call results was an insufficient cffDNA fraction (defined as a fetal fraction of < 4% in 9 studies). In one study, it was defined as < 2.8%, and the median fetal fraction in the no-call group was significantly lower (3.1%) than in the group with a conclusive result (9.1%).
In addition to an insufficient cffDNA fraction, other reasons cited for the no-call results were: unusually high total free DNA, failure to pass the quality control measures and sequencing run failure.
In most cases, the sample was successfully rerun, giving a valid result and a much lower final no-call rate compared to that of the first sample run. In one study, 94% of the no-call results were successfully rerun with the same sample, whereas 3 cases required a third rerun to obtain results. Similarly, there were originally 509 samples with a no-call result in a 2020 study, which was reduced to 46 samples after retesting. As for the 2021 study by Luo Y et al ., 38 participants had a definitive no-call, as opposed to 468 participants after the first sample run.
Discussion:
Main findings
The results have demonstrated that NIPT with cffDNA is reliable in the screening of autosomal trisomies. Overall, T21 had the most promising results, whereas T13 had the least promising. NIPT is also useful for SCAs screening, although it has yielded poorer results overall than those for the autosomal trisomies. Further work is required to improve PPV of SCAs, especially for 45,X0. All study authors agreed that confirmation by invasive procedures are necessary in all cases of high-risk cffDNA results.
Interpretation of the results
Aneuploidies
Amongst the autosomal trisomies, T21 had the highest and T13 had the lowest PPVs. Amongst the SCAs, the lowest PPVs were found for 45,X0. For each aneuploidy, the results oscillated greatly between the studies.
It is important to highlight that the patients with high-risk NIPT results who declined further invasive testing could have significantly impacted the resulting PPV, either in-creasing it if they were true positives or decreasing it if they were false positives.
Discordant results
Most of the studies only listed some general causes of false results. Mosaicism can be of maternal or fetal origin, given that as maternal age increases, a natural loss of the X chromosome results in a maternal cell-free DNA with less X fragments, and that for SCAs, many cases can have multiple cell lines. Another vastly mentioned cause of discordant results is the presence of a vanishing twin, which justifies classifying this factor as an exclusion criteria in our study.
A number of studies confirmed the origin of the discordant results: microdeletion, undiagnosed maternal SCA and mosaicism. These are all well-known causes and are in accordance with a prior systematic review, which found mosaicism to be amongst the main reasons of discordant results.
Although the causes of FP were not verified in most studies, as they would require further testing, such as placental investigation, the number of FP cases were detailed, as this value is needed to calculate the PPV.
With regards to FN, they were poorly reported, or not reported at all. No FNs were identified in 5 studies. However, the underestimation of FNs is very likely given that confirming FNs would require a karyotype of the fetus/newborn, following a negative cffDNA test. A phenotypical postnatal follow-up would not be sufficient to discard an aneuploidy, especially for SCAs as they are phenotypically diverse and can be undetectable at birth. In fact, historically, 75-90% of patients with SCAs remain undiagnosed in their lifetime.
Inconclusive results
The main cause for no-call results was an insufficient cffDNA fraction. The cut-off value was different in one study due to the different DNA sequencing technique used (Single nucleotide polymorphisms technology) (see Table S1), which is claimed to provide accurate results with fetal fractions ≥2.8% (as opposed to ≥4% by MPS) and yet the median fraction was of 3.1% in the no-call group .
The no-call results were missing in 5 studies, possibly because an inconclusive result could have been immediately discarded, or successfully automatically rerun. In one study, all the fetal fractions were ≥4% (4-31%), which may explain the absence of no-call results. Consequently, the fetal fraction significantly impacts cffDNA testing.
While in most cases the sample was successfully rerun, a second blood draw was sometimes necessary, contrary to the recommendation of the American College for Medical Genetics and Genomics32.
Interpretation of the results in the context of other evidence
Over the years, although various systematic reviews assessing the usefulness of cffDNA in prenatal screening have been published, not many focused on the value of cffDNA testing for autosomal trisomies and for SCAs, and on their comparison. In concordance with our results, a 2017 systematic review and bivariate meta-analysis concluded that NIPT by cffDNA is an accurate tool in the screening of autosomal trisomies and 45,X0, and that FN and inconclusive results were poorly reported. A 2017 meta-analysis on cffDNA demonstrated that cfDNA screening in singleton pregnancies could detect >99% of fetuses with T21, 98% of those with T18 and 99% of those with T13; however, the number of reported SCA cases was too low to be accurately. Also consistent with our findings, a recent systematic review found that NIPT is a reliable screening test for SCAs.
Strengths and limitations
This systematic review was conducted according to the PRISMA 2020 guidelines. A thorough review of the available articles related to the topic was performed on 3 databases.
Compared with a 2017 systematic review and meta-analysis, in which prior studies were included, other factors related to cffDNA were investigated and a meta-analysis of the results was conducted. However, the PPV, a relevant indicator of accuracy, was not reported. Furthermore, of all the SCAs, only 45,X0 was investigated. The reason for the fewer studies in our systematic review is due to the focus on aneuploidies, excluding all the other uses of cffDNA.
This review has some limitations. Firstly, PPV was the only metric used and therefore presents its own biases. Secondly, more larger scale studies are needed. Thirdly, given that pregnancy outcome was not an objective of this review, this aspect was not considered. In addition, the differences in the DNA sequencing techniques used in each study were not considered. Finally, there were publications that were not retrieved, despite our best attempts to contact the authors.
Implications for clinical practice
The findings of this review cautiously support the use of cffDNA testing as a screening test for T21, T18, T13, 45,X0, 47,XXY, 47,XXX, and 47,XYY. However, despite the high PPV, we do not recommend its use as a diagnostic test, and should be followed by an invasive confirmatory diagnostic technique.