Prenatal Genome-Wide Sequencing analysis (Exome or Genome) in detecting pathogenic Single Nucleotide Variants in fetal Central Nervous System Anomalies: systematic review and meta-analysis

Abstract

Prenatal Exome (pES) or Genome (pGS) Sequencing analysis showed a significant incremental diagnostic yield over karyotype and chromosomal microarray analysis (CMA) in fetal structural anomalies. Optimized indications and detection rates in different fetal anomalies are still under investigation. The aim of this study was to assess the incremental diagnostic yield in prenatally diagnosed Central Nervous System (CNS) anomalies. A systematic review on antenatal CNS anomalies was performed according to PRISMA guidelines, including n = 12 paper, accounting for 428 fetuses. Results were pooled in a meta‐analysis fitting a logistic random mixed-effect model. The effect of interest was the incremental diagnostic rate of pES over karyotype/CMA in detecting likely pathogenic/pathogenic Single Nucleotide Variants (SNVs). A further meta-analysis adding the available pGS studies (including diagnostic coding SNVs only) and submeta-analysis on three CNS subcategories were also performed. The pooled incremental diagnostic yield estimate of pES studies was 38% (95% C.I.: [29%;47%]) and 36% (95% C.I.: [28%;45%]) when including diagnostic SNVs of pGS studies. The point estimate of the effect resulted 22% (95% C.I.: [15%;31%]) in apparently isolated anomalies, 33% (95% C.I.: [22%;46%]) in CNS-only related anomalies (≥1) and 46% (95% C.I.: [38%;55%]) in non-isolated anomalies (either ≥ 2 anomalies in CNS, or CNS and extra-CNS). Meta-analysis showed a substantial diagnostic improvement in performing Prenatal Genome-Wide Sequencing analysis (Exome or Genome) over karyotype and CMA in CNS anomalies.

Introduction

Genome‐wide DNA sequencing analysis, which includes both Exome Sequencing (ES) and Genome Sequencing (GS), focuses on finding disease‐causing variants in the genome. ES explores the coding sequence of human genes (1.5–2% of the genome), and is a well‐established tool in the diagnosis of pediatric and adult genetic disease [1, 2].

GS refers to the unbiased sequencing of the genome, without targeted capture, having the potential to detect coding Single Nucleotide Variants (SNVs), Copy Number Variants (CNVs), structural rearrangements while also exploring non-coding sequences such as intronic variants and regulatory region variants. In addition, repeat expansions and mitochondrial DNA (mtDNA) can also be investigated [1, 2].

Currently, ES is extensively performed in post-natal practice, and some professional societies support its use even as a first‐line test in children and adults with developmental and intellectual disabilities [3]. A recent meta-analysis from the American College of Medical Genetics And Genomics (ACMG) Pediatric Exome/Genome Sequencing Evidence-Based Guideline Work Group, on pediatric application of genome-wide sequence analysis, evidenced a diagnostic yield of 38% [3], compared with the diagnostic yield of 21% for CMA and targeted sequence analysis (either single gene or gene panel) [3].

Diagnostic prenatal ES (pES), initially performed as a second- or third-line test in highly selected cases, is presently the predominant genome-wide sequencing approach in prenatal diagnosis [4, 5] Howerer, proper indications are still debated [4, 5].

More recently, prenatal genome sequencing (pGS) studies are emerging and there are ongoing trials evaluating its application as a comprehensive genomic testing disclosing both sequence and structural variants [6]. Given this variability, the diagnostic yield for pES/pGS still varies widely among studies, but sufficient data are available to begin differentiating diagnostic yields by specific organ system or number of organ systems affected [5].

With this aim, we performed a systematic review and meta-analysis on fetuses affected by apparently isolated and non-isolated Central Nervous System (CNS) anomalies, undergoing pES or pGS, focusing the analysis on coding SNVs detection rate only.

Materials and methods

This protocol was registered with the International Platform of Registered Systematic Review and Meta-Analysis Protocols (INPLASY) on 01 May 2023 (registration number INPLASY202350001, publicly available).

Systematic review

The research was conducted following PRISMA guidelines [7]. We searched the Pubmed database (https://pubmed.ncbi.nlm.nih.gov/,) lastly accessed on 31 January 2023 for (“fetus” OR “fetuses” OR “foetus” OR “foetuses” OR “fetal” OR “foetal” OR “prenatal” OR “pre-natal”) AND (“Central nervous system” OR “CNS” OR “brain” OR “cerebral” OR “cerebellar” OR “cerebellum” OR “vermis” OR “vermian” OR “blake” OR “Blake’s” OR “hemispheres” OR “hemispheric” “hemisphere” OR “interhemispheric” OR “posterior fossa” OR “cisterna magna” OR “MCM” OR “Dandy-Walker” OR “Dandy Walker” OR “DWM” OR “hydrocephaly” OR “hydrocephalus” OR “ventriculomegaly” OR “corpus callosum” OR “callosal” OR “ACC” OR “pACC” OR “DCC” OR “Probst” OR “septo-optic dysplasia” OR “SOD” OR “cavum” or “CSP” OR “chiari” OR “acrania” OR “spina bifida” OR “anencephaly” OR “anencephalia” OR “ anencephalic” OR “hydranencephaly” OR “hydranencephalia” OR “schizencephaly” OR “schizencephalic” OR “porencephaly” OR “porencephalic” OR “cephalocele” OR “encephalocele” OR “meningocele” OR “meningoencephalocele” OR “neural tube” OR “cerebrospinal fluid” OR “spinal fluid” OR “CSF” OR “NTD” OR “microcephaly” OR “megalencephaly” OR “hemimegalencephaly” OR “holoprosencephaly” OR “HPE” OR “cortical” or “cortex” OR “sulcus” OR “sulci” OR “fissure” OR “fissures” OR “gyrus” OR “gyri” OR “gyra” OR “subcortical” OR “lissencephaly” OR “cobblestone” OR “pachygiria” OR “polymicrogyria” OR “agyria” OR “heterotopia” OR “telencephalon” OR “telencephalic” OR “prosencephalon” OR “prosencephalic” OR “diencephalon” OR “diencephalic” OR “brainstem” or “brain stem” OR “mesencephalon” OR “mesencephalic” OR “pons” OR “pontine” OR “pontocerebellar” OR “medulla” OR “medullar” OR “arachnoid” OR “dural” OR “neuronal migration” OR “migrational” OR “encephalomalacia“ OR “rhombencephalosynapsis” OR “grey matter” OR “white matter” OR “periventricular” OR “encephalopathy” OR “encephalopaties” OR “leukoencephalopathy” OR “acqueduct” OR “ependymal” OR “ependyma”) AND (“WES” OR “CES” OR “exome sequencing” OR “Mendeliome” OR “genome sequencing” OR “GS” OR “WGS” OR “Whole-exome” OR “Whole-genome” OR “medical-exome” OR “clinical-exome”) with a 10-year filter for publication date. A total of 486 titles and abstracts were examined. Only papers with full text available in English language were retained. The risk-of-bias analysis was carried out with the Joanna Briggs Institute (JBI) Critical Appraisal Tool for Case Series [8]. The papers were evaluated by three operators (EM, DG, GM), first independently, then conjointly. Studies involving only postnatal cases were excluded. Case reports and papers describing less than 5 cases were excluded, as small samples are less expressive for estimating the incremental diagnostic yield of genome-wide sequencing analysis over CMA/Karyotype and can leverage too much on the meta-analysis results. Papers not providing quantitative data on pES or pGS in cohorts of fetuses with CNS anomalies were excluded. Studies including fetuses selected based on the suspicion of a specific genetic disorder class were excluded. Papers describing pES or pGS performed after negative targeted panels were excluded. Papers with recurrent phenotypes as an explicit selection criterion were excluded. The exclusion criteria were applied to ensure homogeneity in the enrolling criteria for included papers, minimizing selection bias that could skew of the diagnostic yield, such as fetal demises, Terminations Of Pregnancy (TOPs) performed before genetic testing, phenotype recurrence, previous negative gene panels and reporting bias (individual case reports usually include only clinically significant findings as opposed to systematic cohorts). The geographic origin of the cohorts, the primary indication for genetic testing and the prenatal molecular diagnostic approach adopted were annotated for each study. The cases were divided in the following categories: apparently isolated CNS anomaly (± minor anomalies/soft markers), non-isolated (≥ 2 anomalies, at least one being CNS), CNS-only (≥1 CNS anomalies without multisystemic involvement), multisystemic (malformations involving CNS and any other system).

The incremental diagnostic yield of pES or pGS over karyotyping and CMA was scored for each study and for each category, including only pathogenic (P) and likely pathogenic (LP) coding SNVs in the count. CNVs and SVs analyses were not included in this meta-analysis. When available, the following data were also collected and annotated: Turn-Around-Time (TAT) of the test, the number of Variants of Uncertain Significance (VUSs), secondary and incidental findings (SFs/IFs), the rates of pregnancy terminations and the number of cases in which molecular results had a significant impact on parental decision-making and pregnancy management.

Meta-analysis

The effect of interest was the incremental diagnostic rate of pES over karyotype and CMA. We considered the diagnostic yield of pES over the number of non-diagnosed fetuses through karyotype and CMA in all N = 10 studies eligible for meta-analysis and pooled the results [9–18].

To obtain the widest number of fetus representations, we performed a further meta-analysis adding the two available pGS studies (with ≥ 20–30X depth of coverage) in literature [19, 20]. Since the effect of interest of this meta-analysis is the incremental diagnostic rate of pES over karyotype/CMA, we included only pES-detectable coding SNVs as diagnoses, excluding diagnostic CNVs and SVs from the count. A total of N = 12 articles were included (N = 10 pES and N = 2 pGS). Lastly, we performed a meta-analysis considering three different subcategories: apparently isolated CNS anomaly, non-isolated CNS anomalies (two or more CNS anomalies, or CNS and extra-CNS anomalies), CNS-only related anomalies (one or more). The categories for the possible phenotype association presentation are non-mutually exclusive and partially overlapping, to explore all the available classifications and associations of the index CNS anomaly in literature. We fitted a logistic random mixed-effects model with intercept only. The choice of a mixed-effect model, with a random component, was driven by the need of accounting for between-study heterogeneity. 95%-Clopper-Pearson confidence intervals (C.I.) for individual studies were calculated. The between-study variance (Tau-squared) was estimated through maximum likelihood estimator. The results are illustrated in Forest plots. In the second model, between-study heterogeneity needed further investigation. A leave-one-out analysis was performed and a Baujat plot was included [21]. All statistical analyses were performed in R 4.2.2 (R Core Team 2022) using the meta package v6.1-0 [22, 23].

Results

Systematic literature review

The review results are presented in Fig. 1 according to the PRISMA guidelines [7]. N = 486 papers were identified from the initial Pubmed search. After title and abstract screening, N = 464 were excluded as they did not feature prenatal cohorts of fetuses undergoing pES during pregnancies and selected exclusively for CNS anomalies. N = 22 papers were sought for retrieval, and the full text was available for all of them. After full-text examination, N = 11 were excluded from quantitative analysis. Specifically, N = 4 were excluded as they did not meet the requirements for the number of analyzed cases, N = 2 were excluded as they did not provide quantitative diagnostic parameters, N = 2 were excluded as the enrollment criteria focused on the suspicion of specific genetic syndromes which may or may not have CNS anomalies, N = 2 were excluded as they only featured postnatal cases, and N = 1 paper was excluded as it involved only familial/recurrent cases. The detailed risk-of-bias analysis for the excluded papers is provided in Supplementary Material 1 - List of papers excluded after full-text examination.

Fig. 1. PRISMA Flow Chart.

The figure illustrates the systematic review process according to the PRISMA guidelines [7].

The remaining N = 11 papers were included for quantitative analysis [9–16, 18–20]. N = 1 paper was retrieved from citation searching from the included studies, and added to quantitative analysis [17], resulting in the final number of N = 12 papers reporting cohorts of fetuses undergoing pES or pGS for CNS anomalies, after non-diagnostic karyotype and CMA results. Information on the geographic origin of the cohorts, the specific test performed, diagnostic yields for each category (apparently isolated, non-isolated, CNS-only, multisystemic), TAT, the rates of VUSs, IFs/SFs, TOPs and clinical impact of the results is presented in Table 1.

Table 1.

Results from studies eligible for quantitative analysis.

				DIAGNOSTIC YIELD
Source	Area	Primary Indication	Test	TOTAL	APPARENTLY ISOLATED CNS anomaly ± minor anomaly	NON ISOLATED >= 2 anomalies (CNS or CNS + extra CNS)	CNS-ONLY one or more CNS anomalies	MULTISYSTEMIC Involving CNS and other systems	TAT (days)	VUS	IFs and/or SFs	TOP	Clinical Impact
Lei et al. [10]	China	Corpus callosum anomalies	trio-ES	17/50 (34.00%)	10/34 (29.41%)	7/16 (43.75%)	n.a.	n.a	35	n.a.	n.a.	34/49 (69.39%)	17/50 (34.00%)
Yaron et al. [9]	Israel	CNS anomalies	quatro-WES 2 trio-WES 82 duo-WES 1 singleton-WES 1	38/86 (44.18%)	10/30 (33.33%)	28/56 (50.00%)	24/54 (46.30%)	14/32 (43.75%)	n.a.	9	n.a.	n.a.	n.a.
Yang et al. [20]	China	CNS anomalies	WGS	24/127 (18.90%)	11/82 (13.41%)	13/45 (28.89%)	12/101 (11.88%)	12/26 (46.15%)	21	n.a.	3 IFs 3 SFs	n.a.	n.a.
Liao et al. [19]	China	abnormal sylvian fissure	WGS	12/24 (50.00%)	n.a.	12/24 (50.00%)	6/15 (40.00%)	6/9 (66.67%)	n.a.	n.a.	n.a.	10/24 (41.67%)	10/12 (83.33%)
de Koning et al. 2021 [11]	Netherlands	CNS anomalies	trio-WES	10/19 (52.63%)	1/3 (33.43%)	9/16 (56.25%)	2/8 (25.00%)	8/11 (72.73%)	17.35	2	n.a.	7/19 (36.84%)	12/19 (63.16%)
She et al. [12]	China	Corpus callosum anomalies	trio-WES	2/5 (40.00%)	2/5 (40%)	n.a.	2/5 (40.00%)	n.a.	n.a.	n.a.	n.a.	5/5 (100%)	2/5 (40.00%)
Heide et al. [13]	France	Corpus callosum anomalies	trio-WES	12/62 (19.35%)	6/44 (13.64%)	6/18 (33.33%)	8/n.a.	4/n.a.	21.5	6	n.a.	9/62 (14.52%)	53/64 (82.81%)
Tan et al. 2020 [14]	China	CNS anomalies	trio-CES	5/11 (45.45%)	1/5 (20.00%)	4/6 (66.67%)	4/8 (50.00%)	1/3 (33.43%)	n.a.	n.a.	n.a.	5/11 (45.45%)	3/5 (60.00%)
Li et al. [15]	China	cerebellar vermis anomalies	trio-WES 11 singleton WES 8	8/19 (41.11%)	2/7 (28.57%)	6/12 (50.00%)	2/7 (28.57%)	6/12 (50.00%)	42	2	n.a.	10/19 (52.63%)	5/19 (26.32%)
Weitensteiner et al. [16]	Germany	CNS anomalies	trio-WES	3/6 (50.00%)	0/1 (0.00%)	3/5 (60.00%)	2/4 (50.00%)	1/2 (50.00%)	n.a.	1	n.a.	n.a.	n.a.
Reches et al. [17]	Israel	CNS anomalies	trio-WES	4/7 (57.14%)	0/2 (0.00%)	4/5 (80.00%)	3/5 (60.00%)	1/2 (50.00%)	n.a.	1	n.a.	6/7 (85.71%)	n.a.
Poirier et al. [18]	France	Microlissencephaly	singleton-WES	3/12 (25.00%)	n.a.	n.a.	3/12 (25.00%)	n.a.	n.a.	n.a.	n.a.	n.a.	n.a.

The table summarizes, when available from the source, the sequencing method applied: Whole Exome Sequencing (WES), Whole Genome Sequencing (WGS), Clinical Exome Sequencing (CES) and the results on diagnostic yield (with subcategories), Turn-around time (TAT), Variants of Uncertain Significance (VUS), Incidental/Secondary findings (IFs/SFs), Terminations of Pregnancy (TOP), and clinical impact of the results from papers included in quantitative analysis.

The risk-of-bias analysis (JBI Critical Appraisal Checklist for Case Series) of the included papers is provided in Supplementary Material 2. The critical appraisal showed an overall low risk-of-bias for included studies. However, some degree of reporting bias (preferential publication of series with high diagnostic yield) and of selection bias should be taken into account for at least part of the studies. A form of selection bias occurs in studies in which not all eligible cases might have undergone pES/pGS. This might be due to study design in retrospective series studies [14, 16], or due to self-selection bias. Self-selection bias occurs when couples refuse to undergo the specific offered testing. This might be due to clinical reasons alone [9, 10, 15], but in some cases economic factors are also taken into account [17].

Meta-analysis

Meta-analysis of pES and pES/pGS incremental diagnostic yield

Firstly, we calculated the point estimate of the pooled incremental diagnostic yield of pES for SNVs over CMA/karyotype in isolated and non-isolated CNS anomalies. The between study-heterogeneity was adequate, as suggested by the value of Thompson’s I² = 38% statistic and a non-significant p-value of the Wald-Type test on Cochran’s Q (p = 0.10) (Fig. 2, panel A).

Fig. 2. Meta-analysis - Incremental diagnostic yield of prenatal genome-wide sequencing in all fetuses with CNS anomalies (either isolated or associated, single or multiple).

The figure illustrates the point estimate of the incremental diagnostic yield of prenatal genome-wide sequencing in CNS anomalies (isolated and/or associated) from the papers included in the quantitative analysis. The figure also presents the number of cases (Total), diagnoses (Events), the diagnostic yield for each study with 95% C.I. The between study-heterogeneity is assessed with Thompson’s I² and Wald-Type test on Cochran’s Q. The results are presented in table format and as a Forest plot. Panel A: meta-analysis including only pES studies. Panel B: meta-analysis including pES studies and pES-detectable diagnostic SNVs from pGS studies.

The pooled incremental diagnostic yield estimate was 38% (95% C.I.: [29%;47%])_, showing a substantial gain over CMA/karyotype alone in CNS fetal anomalies (Fig. 2, panel A).

Secondly, we calculated the point estimate of the pooled incremental diagnostic yield of prenatal genome-wide sequencing analyses (including only diagnostic SNVs from pES and coding SNVs from pGS studies) over CMA/karyotype in CNS anomalies. The pooled estimate of the effect was 36% (95% C.I.: [28%;45%]) (Fig. 2, panel B). The meta-analysis including pGS shows a substantial gain with respect to performing CMA/karyotype alone. However, a high between-study heterogeneity was detected both by a high value of Thompson’s I² = 65% statistic and a significant p-value of the Wald-Type test on Cochran’s Q (p < 0.01) (Fig. 2, panel B).

To better analyze this heterogeneity, an influence analysis of the different studies was performed. The Baujat plot was inspected (Fig. 3, panel A) [21], observing that the studies by Yaron et al. [9] and Yang et al. [20] have a very high leveraging on the pooled result (Fig. 3, panel A) [9, 20].

Fig. 3. Influence analysis and leave-one-out analysis.

Panel A: Baujat plot [21]. The plot illustrates the influence analysis and the overall heterogeneity contribution of pES studies and pGS studies included in the meta-analysis. Panel B: The figure illustrates the point estimate of the incremental diagnostic yield of prenatal genome-wide sequencing in CNS anomalies (isolated and/or associated) from the papers included in the quantitative analysis leaving out the study of Yang et al. [20]. The figure also presents the number of cases (Total), diagnoses (Events), the diagnostic yield for each study with 95% C.I. The between study-heterogeneity is assessed with Thompson’s I² and Wald-Type test on Cochran’s Q. The results are presented in table format and as a Forest plot.

In particular, Yaron et al. [9] has a high weight on the estimate [9], but it does not contribute substantially to heterogeneity, as the corresponding cohort is large compared to other studies, but the single-study conclusion is in line with the others. Conversely, Yang et al. [20] represents the true outlier among all the studies [20]. In this case, as in Yaron et al. [9], the influence on the pooled estimate is high, reporting the largest cohort, but the contribution to between-study heterogeneity is remarkable. This can be attributed to the lower estimated diagnostic yield in the study.

To obtain a more precise insight on the leveraging of single studies beyond the visual Baujat plot inspection (Fig. 3, panel A), a leave-one-out analysis was performed.

We fitted the model again, leaving out one study at a time. As expected, leaving out the study by Yang et al. [20] reduces the heterogeneity (I² = 37%) and the p-value of the Wald-type test on Cochran’s Q is non-significant (p = 0.10), whereas the effect does not change substantially (diagnostic yield 39%) (Fig. 3, panel B).

We conclude that this corroborates the validity of the result, confirming that there is a diagnostic improvement in performing genome-wide sequencing analyses.

The inclusion of the two pGS studies in the meta-analysis did not influence the estimated yield. We remind that, for pGS, we only included coding (pES-detectable) SNVs. Indeed, the meta-analysis considering only pES studies gives SNVs detection results comparable to the meta-analysis excluding the outlying study by Yang et al. [20]: the point estimates are almost overlapping and the 95% C.I.s are overlapping. Considering this result, the two pGS studies were included in the following meta-analysis, analyzing three subcategories: apparently isolated CNS anomalies, non-isolated anomalies (two or more CNS anomalies, or CNS and extra-CNS), and CNS-only related anomalies (one or more).

Notwithstanding this observation, we underline the representation pGS studies is still scarce and pGS has the potential to incorporate all diagnostic analyses in a single test, including also SVs, CNVs, intronic variants, regulatory region variants, mtDNA variants and repeat expansions, not investigated in this meta-analysis.

Further data must be collected, and our models have to be updated consequently.

Meta-analysis on subcategories of pES/pGS incremental diagnostic yield

We calculated the point estimate of the pooled incremental diagnostic yield of prenatal genome-wide sequencing (SNVs from pES and pGS) in fetuses presenting apparently isolated CNS anomalies. In this first subcategory, the two studies of Weitensteiner et al. [16] and Reches et al. [17], displayed no incremental diagnostic yield (Table 1) [16, 17]. However, the total number of fetuses presenting an apparently isolated CNS anomaly was respectively one and two.

The sample number is too low to be representative, and this substantially increases the uncertainty of the conclusion that there is no gain in performing pES analysis in those cases. The two studies were hence excluded from the sub-meta-analysis. The pooled incremental diagnostic yield estimate was 22% (95% C.I.: [15%;31%]) (Fig. 4, panel A), showing a substantial gain compared to CMA/karyotype alone even in presence of an apparently isolated CNS anomaly.

Fig. 4. Subcategory Meta-analysis Forest Plots - incremental diagnostic yield of pES/pGS in fetuses with an apparently isolated CNS anomaly, non-isolated CNS anomaly/anomalies or CNS-only anomalies (1 or more).

The figure illustrates the point estimate of the incremental diagnostic yield of prenatal genome-wide sequencing in three subcategories of fetuses with CNS anomalies. The figure also presents the number of cases (Total), diagnoses (Events), the diagnostic yield for each study with 95% C.I. The between study-heterogeneity is assessed with Thompson’s I² and Wald-Type test on Cochran’s Q. Panel A: Fetuses with an apparently isolated CNS structural anomaly. Panel B: Fetuses with non-isolated CNS anomaly/anomalies (≥ 2 CNS or CNS + extra-CNS). Panel C: Fetuses with CNS- only, single or multiple anomalies.

We then calculated the point estimate of the pooled incremental diagnostic yield (SNVs from pES and pGS), in fetuses presenting non-isolated CNS anomaly (either ≥ 2 anomalies in CNS, or CNS and extra-CNS), when available from source papers. The between study-heterogeneity was low, as suggested by the value of Thompson’s I² = 16% statistic (Fig. 4, panel B). The pooled incremental diagnostic yield estimate was 46% (95% C.I.: [38%;55%]), as illustrated in the Forest plot (Fig. 4, panel B), showing a substantial gain compared to performing CMA/karyotype alone, as expected. Lastly, we calculated the point estimate of the pooled incremental diagnostic yield of prenatal genome-wide sequencing (SNVs from pES and pGS) in the subcategory of fetuses presenting CNS-only related anomalies (≥1) when available from source papers. The between-study heterogeneity was high, as suggested by the value of Thompson’s I² = 63% statistic and a significant p-value of the Wald-Type test on Cochran’s Q (p < 0.01) (Fig. 4, panel C). However, we do not believe this can represent a limitation to the validity of the study, since we have good evidence from the complete meta-analysis that there is a gain in performing prenatal genome-wide sequencing studies in fetuses presenting CNS anomalies. The pooled incremental diagnostic yield estimate was 33% (95% C.I.: [22%;46%]), as illustrated in the Forest plot (Fig. 4, panel C). We conclude that the meta-analysis shows a substantial gain with respect to performing CMA/karyotype alone in presence of one or more than one CNS anomalies, either associated to other CNS anomalies, or to other multisystem anomalies.

Discussion

In the last five years, pES has been increasingly adopted in the clinical prenatal setting, usually focusing the analysis on a specific pool of selected genes (i.e. in silico panel) based on the clinical indication. In 2019, two large-scale sequencing projects (from cohorts selected between 2014 and 2017) were performed using trio-pES in 610 and 234 fetuses with a wide range of fetal structural anomalies, after the exclusion of aneuploidies/CNVs [24, 25]. These studies reported an incremental diagnostic yield of 8.5% and 10% over karyotype and CMA [24, 25], and paved the way for the widespread use of pES. A recent systematic literature review showed that pES offered for ultrasound (US) fetal anomalies an overall diagnostic yield of 31% over karyotype and CMA [26]. The diagnostic yield in unselected cohorts of fetuses with US anomalies resulted in 15%, but increased to 42% in cases selected for a higher suspicion of a monogenic disorder based on the phenotypic presentation and/or family history [26].

Despite encouraging literature data, more homogeneity is needed in establishing access criteria.

The first inter-society position statement providing a framework for pES discouraged routine use of pES [27], proposing its use mainly in research, with prompts for single-case evaluation. In 2020, the ACMG published a statement suggesting that pES might be considered after non-diagnostic results from routine prenatal testing in fetuses with one or more significant anomalies [28].

The last inter-society position, published in April 2022 [5], recognized that recent fetal diagnostic sequencing application provides sufficient experience to allow clinical use. The statement concludes that literature data support prenatal sequencing in fetuses with major single or multiple anomalies, specifying that as pES is not currently validated to detect all CNVs, CMA should be run before or in parallel with pES in this scenario. Further studies are needed to support the decision between a sequential or concurrent application of CMA and pES for different fetal anomalies.

Previous studies in literature reported different diagnostic yields in the CNS subgroups retrieved from cohorts of all kinds of unselected US anomalies. In the cohort analyzed in Petrovski et al., 49 cases belonged to the CNS subgroup, with a diagnostic yield of 22% (11/49) [25], whereas in the PAGE study, Lord et al. reported a diagnostic yield of 4.3% in 69 fetuses with brain anomalies [24]. In literature, a meta-analysis focused on the Agenesis of Corpus Callosum (ACC), the most frequent antenatal diagnosed CNS malformation, has been recently published [29] showing an incremental diagnostic yield of pES over CMA of 55% (95% C.I.: [35%; 73%]) in ACC with extracranial anomalies, 43% (95% C.I.: [30%; 57%]) in ACC with other cranial anomalies and 32% (95% C.I.: [18%; 51%]) in (apparently) isolated ACC [29].

The aim of this work was to assess the incremental diagnostic yield of genome-wide sequencing analysis (focusing on coding SNVs only) after karyotype and CMA inconclusive results in cohorts of fetuses selected for CNS anomalies. We performed a systematic review of the literature following PRISMA guidelines and the results were pooled in a meta-analysis [7]. We included only papers focused on CNS anomalies, ruling-out specific phenotypic subgroups from previously published larger cohorts focused on all kind of fetal anomalies (i.e. Lord et al. and Petrovski et al. [24, 25]) to avoid double-counting of cases.

A total of 12 articles, 10 investigating pES and 2 pGS, accounting for 428 fetuses, were included in the systematic review [9–20]. The diagnostic yield of prenatal genome-wide sequencing methods (considering only diagnostic SNVs detected either by pES or pGS) in isolated and non-isolated CNS anomalies, ranged from 18.90 to 57.14% (Table 1).

The pooled incremental diagnostic yield estimate of pES over CMA/karyotype in CNS anomalies was 38% (95% C.I.: [29%;47%]) (Fig. 2, panel A), showing a substantial gain with respect to CMA/karyotype alone. We performed a further meta-analysis also including the two pGS studies, including only the diagnostic SNVs in the diagnostic yield. The results were confirmed, with a pooled estimate of the effect of 36% (95% C.I.: [28%;45%]) (Fig. 2, panel B), corroborating the validity of performing genome-wide sequencing analysis in fetal CNS anomalies.

Notwithstanding this observation, we underline that the representation of pGS studies is still scarce in literature and further data must be collected to update models and meta-analysis, as the potential greater diagnostic yield of pGS concerns also CNVs, SVs, and intronic variants, not investigated in this meta-analysis.

We also calculated the point estimate of the pooled incremental diagnostic yield of prenatal genome-wide sequencing for SNVs in three subcategories. For fetuses presenting only apparently isolated CNS anomalies the incremental diagnostic yield estimate was 22% (95% C.I.: [15%;31%]) (Fig. 4, panel A). For fetuses presenting non-isolated CNS anomaly (either ≥ 2 anomalies in CNS, or CNS and extra-CNS), the pooled incremental diagnostic yield estimate was 46% (95% C.I.: [38%;55%]) (Fig. 4, panel B), as expected higher than in apparently isolated CNS anomalies. Lastly, in fetuses presenting CNS-only related anomalies (≥1), the pooled incremental diagnostic yield estimate was 33% (95% C.I.: [22%;46%]) (Fig. 4, panel C), a result in-between the incremental diagnostic yield of apparently isolated CNS anomalies and non-isolated CNS anomalies. In conclusion, the meta-analysis showed a substantial diagnostic improvement for the application of genome-wide sequencing analyses over karyotype and CMA alone, as SNVs appear to be a large proportion of prenatal causes of CNS malformations in all subcategories of CNS anomalies, whether apparently isolated, associated with other CNS anomalies, or with other multisystem anomalies.

Notwithstanding this observation, it should be underlined that reported diagnostic yields must be tempered by several factors including the sequencing approach, the number of genes analyzed (especially if limited to a chosen subset of genes from the original sequencing experiment), and the laboratories’ practices on defining pathogenic variants. Unfortunately, such information is not always declared in the studies, and a formal analysis of the impact of these variables was not possible.

Benefits and challenges

There is limited evidence on the neurodevelopmental outcomes of children affected by CNS anomalies diagnosed prenatally. Predicting the risk of developmental difficulties is challenging. An integrated approach with neuroimaging, infectious, and genetic analyses may add useful information [30]. Not all this information may be available at the time of first counseling, which should be a step-wise approach, with further discussions held as more information is obtained. With this regard, pES can add valuable prognostic information for the parental decision-making process, guiding clinical management during pregnancy and the perinatal period [30].

For families who might consider pregnancy termination as a management option on the basis of genetics results, having a short result TAT is needed to respect gestational age limitations set by the law where they reside.

The present review showed that, the median TAT (available in 5/12 studies; 42%) was 27.37 days (range 17.35–42 days) and, in papers describing the clinical impact of pES (n = 7/12, 58%), the results influenced the decision-making process in more than half of cases (55.66%, range 26.32–83.33%) (Table 1).

Phenotypes

Patient formal phenotyping is crucial in defining candidate genes in genome-wide sequencing approaches such as ES and GS. However, prenatal phenotypes of genetic diseases at present are not completely described, may evolve during pregnancy and may be specific to the developmental stage [31]. New phenotypes or signs can be observed in more advanced gestational ages, and phenotypes can resolve (i.e. cystic hygroma or hydrops fetalis) even in the presence of an underlying genetic disease. During routine third-trimester US, an incidental fetal anomaly can be found in about 1 in 300 scanned women. These include anomalies missed at the earlier US scans but also abnormalities that can only be seen with fetal maturation, as with many CNS anomalies, such as anomalies of cortical development, microcephaly, or hydrocephalus [31]. Human Phenotype Ontology (HPO) descriptors are widely used for formal phenotyping to standardize clinical information, representing a comprehensive database of abnormal human phenotypic features including signs, symptoms, laboratory test results, and imaging findings. In addition, the Global Alliance for Genomics and Health is developing a suite of coordinated standards for genomics for healthcare, including the “Phenopacket”, created for sharing disease and phenotype information. For instance, the age of onset of each feature is provided, and the fact that a particular sign was excluded on prenatal US examination is indicated. This effort has the aim to understand the genetic and phenotypic architecture of fetal anomalies and stillbirth and use this information to improve pre- and perinatal, postnatal, and maternal care. As a first application of Phenopacket schema for prenatal medicine, a collaboration between the HPO team and the Fetal Sequencing Consortium was recently started [32].

Other results (VUS, IFs, SFs)

Broad genome-wide sequence analyses, while increasing the predictive potential, will identify a larger number of findings with difficult clinical management, namely VUSs, IFs and SFs. In our quantitative analysis, only one paper reported SFs/IFs [20] and only six papers reported inconclusive results [9, 11, 13, 15–17], explicitly declaring the number of VUSs. We could not score a specific rate due to the lack of homogeneity and the low number of papers reporting these findings, reflecting the managing difficulties for such results in prenatal diagnosis (Table 1). Despite the absence of formal statistical data, we will provide a brief discussion on the state-of-the-art concerning these findings.

VUSs might be identified in known genes as well as genes previously not associated with specific phenotypes. In the last Points to consider document [28], ACMG highlighted how laboratories have different policies regarding prenatal VUSs report, even within the same country. ACMG suggests reporting VUSs in phenotype-fitting genes, especially for autosomal recessive conditions if a VUS is found in trans with a pathogenic/likely pathogenic variant [28]. The last Updated Position Statement of ISPD on prenatal genome‐wide sequencing [5], confirms this approach, underlying the importance of expert genetic post‐test counseling [5].

In a recent single-laboratory experience review [33], authors stated that in most cases, only Pathogenic or Likely Pathogenic variants explaining the fetal phenotype were reported. However, in selected cases, VUSs which may explain the phenotype and only require a single piece of evidence to be upgraded, the so called “hot” class 3s, might be reported to highlight the need for further prenatal surveillance and postnatal investigations if indicated [33], after appropriate pathogenicity variant assessment by a multidisciplinary team (molecular biologists, clinical geneticists, maternal/fetal medicine experts, imagers etc.) [32].

IFs and SFs include variants with possible implications in monogenic conditions unrelated to the original indication [34]. IFs are unexpected results encountered in the analysis, while SFs occur in gene purposely and possibly systematically investigated regardless of the original indication [34].

For postnatal ES and GS, the ACMG recommends reporting P/LP variants in 81 actionable genes causing highly penetrant disorders for which interventions preventing/reducing morbidity and mortality are available, as SFs [34]. This document does not provide specific indications in prenatal diagnosis, however updates are regularly available [34]. The last ACMG statement on prenatal sequencing recommends the opt-out option, only reporting SFs if parents consented [28]. The last ISPD Position Statement confirms the opt-out option, recommending a separate consent for each parent [5].

Interestingly, a recent survey on genetics professionals’ attitudes towards pES highlighted that for most participants the definition of actionable results is different in the prenatal setting due to the option for pregnancy termination [35]. Currently the ACMG statement recommends reporting as IFs, only high-penetrant pathogenic variants known to cause moderate/severe childhood onset disorders [28]. In trio-based analysis, SFs and IFs may be searched for and detected in the parental as well as in the fetal genotypes. The ACMG recommends laboratories to establish clear policies regarding limiting reports to fetal variants or including parental analysis (irrespective of fetal inheritance) [28].

Additional findings such as non-paternity or consanguinity are also detectable, and parents should be aware before testing is undertaken that these may be identified.

Future perspectives

According to the last ISPD statement, there is currently no evidence supporting pES routine testing (including upon parental request) for indications other than fetal anomalies [5]. However, in literature some papers exploring the possibility of pES use in pregnancies without structural anomalies are available. In 2022, Vaknin et al., performed pES in 210 pregnancies without major fetal ultrasound findings: 6/210 (2.86%) cases were found to have P/LP variants, showing an unexpected rate of diagnoses in pregnancies with no or minor US anomalies [36]. Should these results be confirmed, the question on a clinical pES application in the absence of major fetal anomalies would be open.

Another possible future scenario is the substitution of karyotype/CMA plus pES analysis with pGS analysis. As previously discussed, GS allows the sequencing of an individual’s whole genome, potentially identifying more genetic variations than any other approach, raising further questions about interpretation and reporting of the variants. Two different strategies are currently available. The first one is the high-coverage GS, which explores anomalies ranging from SVs to CNVs and SNVs in a single analysis [6]. The second approach is CNV-focused low-coverage genome sequencing [37]. Strong evidence supports that this approach, identifying gross and submicroscopic CNVs and chromosomal anomalies, could be applied as an alternative to CMA in prenatal diagnosis [37], requiring less DNA (thus reducing the need for cell culture) and performing better in regions with lower CMA probes density.

Concerning GS, there is preliminary prenatal and postnatal data demonstrating comparable diagnostic yields between GS and the association of ES and CMA [6, 37, 38], potentially with lower cost for GS depending on the sequentiality or concurrency of ES and CMA.

In this scenario, early use of genome-wide sequencing could enable more timely diagnosis for patients with unexplained developmental delay or multiple congenital anomalies in postnatal setting [3], and probably in prenatal setting too, opening a possible larger application soon [1].

It should be underlined that, when no potential pathogenic variants are disclosed, the uninformative result can be mistakenly interpreted as a reassuring finding of a normal outcome. A non-diagnostic analysis reduces the risk of a genetically determined condition, however we are still not able to estimate a proper residual risk. Long term follow-up of born children with non-diagnostic results is of utmost importance to provide accurate prenatal counseling when all genetic analyses are inconclusive.

In conclusion, this large overview highlighted current evidence on advantages and disadvantages of pES (and pGS) and future directions. These approaches are revolutionizing prenatal medical genetics and the management of fetal US anomalies.

The meta-analysis showed a substantial diagnostic improvement in performing genome-wide sequencing analysis over standard and molecular cytogenetics, supporting the proposal of offering pES earlier in the diagnostic route of CNS fetal anomalies.

It might be advisable to perform pES in parallel with CMA, bearing in mind that, in the near future, this approach might be replaced by pGS analysis as a first-tier test.

Author contributions

Study conceptualization, EM; data curation EM, DG, GM, formal analysis, EM; funding acquisition N/A; Investigation N/A; methodology, EM; DG; GM; project administration EM, AP; Resources: EM, DG, GM; writing—original draft preparation EM; DG, writing—review and editing EM, DG; GM; AP; manuscript supervision AP.

Funding

The study received no funding.

Data availability

All data will be available upon reasonable request at the corresponding author.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41431-024-01590-2.