Added value of chromosomal microarray analysis over karyotyping in early pregnancy loss: systematic review and meta-analysis
ABSTRACT
Objective
To estimate the increased test success rate and incremental yield of chromosomal microarray analysis (CMA) over conventional karyotyping in detection of pathogenic copy number variants (CNVs) and variants of unknown significance (VOUS) in early pregnancy loss.
Method
This was a systematic review conducted in accordance with PRISMA criteria. All articles identified in PubMed, Ovid MEDLINE and Web of Science, between January 2000 and April 2017, that described CNVs in early pregnancy losses (up to 20 weeks) were included. Risk differences were pooled to estimate the incremental yield of CMA over karyotyping overall, and after stratification. In addition, test success rate, defined as the proportion of informative results, was compared in series in which CMA and karyotyping were performed concurrently.
Results
Twenty-three studies, reporting on 5507 pregnancy losses up to 20 weeks with full data available, met the inclusion criteria for analysis. In the series in which CMA and karyotyping were performed concurrently, CMA showed a significant improvement in success rate, providing informative results in 95% (95% CI, 94–96%) of cases compared with karyotyping in which informative results were provided in 68% (95% CI, 66–70%) of cases. Combined data from reviewed studies revealed that incremental yields of CMA over karyotyping were 2% (95% CI, 1–2%) for pathogenic CNVs and 4% (95% CI, 3–6%) for VOUS. The most common pathogenic CNVs reported were 22q11.21 and 1p36.33 deletion.
Conclusion
In comparison with conventional karyotyping, CMA provides a significant increase in test success rate and incremental diagnostic yield in early pregnancy loss. Copyright © 2017 ISUOG. Published by John Wiley & Sons Ltd.
INTRODUCTION
Early pregnancy loss, commonly known as miscarriage, is the most frequent adverse outcome in early pregnancy, occurring in 10–20% of clinical pregnancies1. Although its upper limit is usually considered to be 20 or 24 weeks, 80% of losses occur during the first 12 weeks of pregnancy2. Chromosomal anomalies are found in more than half of early pregnancy losses, and they are less common thereafter3, 4. On karyotyping, autosomal trisomies are shown to be the most frequent anomalies (65%) associated with first-trimester miscarriage, followed by triploidy (13%) and monosomy X (10%)5, 6. It has been shown that the etiologic analysis of a miscarriage has a major impact on the woman's/couple's future reproductive plans and prenatal care in future pregnancies7.
Karyotyping has been used traditionally for genetic testing in pregnancy loss, but this technique has many limitations, especially with regards to analysis of products of conception. Firstly, the quality and viability of samples are often suboptimal resulting in failed cultures in 20–40% of cases8-10. Secondly, presence of maternal-cell contamination, a common occurrence in products of conception, is undetectable by karyotyping in female fetuses, leading to false-negative results in 29–58% of samples11, 12. Chromosomal microarray analysis (CMA), either using array comparative genomic hybridization (aCGH) or single nucleotide polymorphism (SNP) genotyping, can replace conventional karyotyping due to its less rigorous requirements for sample quality and its higher resolution. CMA is a molecular technique that detects copy number variants (CNVs), submicroscopic gains or losses of DNA, with resolution down to 10 kb within the genome, undetectable by karyotyping. A previous meta-analysis assessed CMA performance in early pregnancy loss and demonstrated a 13% incremental yield of CNVs undetectable by karyotyping13.
In this systematic review, we aimed to estimate the increased testing success rate and incremental yield of CMA over conventional karyotyping in detection of pathogenic CNVs and variants of uncertain significance (VOUS) in early pregnancy losses.
METHODS
Search strategy and eligibility criteria
This review was performed according to PRISMA guidelines for conducting systematic reviews14. A systematic search focused on studies performing CMA on first- and second-trimester pregnancy losses. The search terms were initially agreed and studies were selected in a three-step process. The complete search string is outlined in Table S1. Electronic searches of PubMed, Ovid MEDLINE and ISI Web of Knowledge (Web of Science) databases were performed, using the search terms: fetal or prenatal, and pregnancy loss or miscarriage, and array comparative genomic hybridization or copy number variants, with related search terms, from January 2000 to April 2017. All titles and abstracts were reviewed by two researchers (M.P. and M.G.) and potentially eligible full-text articles were extracted. Differences were resolved by discussion with a third researcher (A.B.).
After screening the available abstracts, studies were selected if CMA was applied to products of conception obtained from miscarriages (up to 20 weeks). In some studies, CMA was performed after a normal karyotype was obtained, in others concurrently with quantitative fluorescent polymerase chain reaction (QF-PCR) or conventional karyotyping, and, in the remaining, CMA was used as standalone test. Third-trimester fetal losses and stillbirths were not considered, nor were studies applying a technique other than CMA (including conventional comparative genomic hybridization, the precursor technique of aCGH). Studies were excluded when data could not be extracted, when gestational age and indication stratification were not provided in series including late losses or other indications for CMA, respectively, and when only morphologically abnormal embryos were included or could not be identified. The full texts of only original research articles in English on CMA of first- and second-trimester pregnancy loss were reviewed. As a review of previously published data, the study did not require ethics approval. The data were not used for any purpose other than those of the original study, and no new data were collected.
Data extraction
Data on pregnancy losses and normal karyotype were identified and included in the analysis. When karyotyping was not performed prior to CMA (either performed concurrently or not performed at all), cases with a genomic imbalance > 10 Mb were considered detectable by karyotyping, and therefore excluded from our analysis. However, those cases were included in the analysis when a previous karyotype overlooked them. Data on inclusion criteria, gestational period, CMA type and resolution, and description of pathogenic CNVs and VOUS, were extracted from these studies. Authors of series on CMA in early pregnancy loss in which results were not stratified by trimester, were contacted and asked to provide further data. Details of all reported CNVs were reviewed by two authors independently (M.P. and L.R.-R.) to evaluate clinical significance and only those deemed pathogenic were considered. Any disagreement was resolved by discussion.
Assessment of risk of bias
Quality assessment of included studies was performed using the Quality Assessment tool for Diagnostic Accuracy Studies (QUADAS-2) checklist, and risk of bias and applicability were assessed according to patient selection, index test, reference standard and flow and timing15. Each parameter was graded as having high, unclear or low risk of bias. The risk of bias was measured individually by two reviewers (M.P. and M.G.). Publication bias was assessed graphically using funnel plots.
Data synthesis
Test success rates, defined as the proportion of informative results, by karyotyping and CMA were noted when reported, and 95% CI for proportions were used to assess statistical significance. The overall increase in test success rate was calculated as a proportion and as a delta value in proportion points between both techniques. Incremental yield (risk difference) of CMA was defined as the yield over karyotyping for each prenatal series. Risk differences were pooled (weighting by inverse variance) with 95% CI to estimate the overall CMA incremental yield, using RevMan version 5.3.4 (Cochrane Informatics and Knowledge Management Department, Copenhagen, Denmark) and the corresponding forest plots were constructed for both pathogenic CNVs and VOUS. Between-study heterogeneity was assessed using the tau2, X2 (Cochrane Q) and Higgins I2 statistics. According to the Cochrane handbook, the heterogeneity measured by I2 is interpreted as non-important (< 30%), moderate (30–60%) or substantial (> 60%)16. A random-effects model was used when there was significant heterogeneity. Analysis was also performed for detection of chromosomal anomalies by conventional karyotyping (when reported), for pathogenic CNVs and for VOUS. Analysis of pathogenic CNVs was stratified according to whether series included recurrent pregnancy losses. Forest plots were constructed for (a) chromosomal anomalies by karyotyping; (b) overall pathogenic CNVs by CMA; (c) pathogenic CNVs by CMA stratified according to recurrent fetal loss; and (d) VOUS. Publication bias was assessed by funnel-plot asymmetry and quantified using Egger's method17, 18.
RESULTS
The search identified 630 articles, 29 of which were full-text reviewed and 20 of those were finally included in the meta-analysis9, 19-37. Nine studies were excluded either because reported data could not be extracted, or identification of the gestational age or indication for sampling for each case was not provided (if late pregnancy losses or other indications for CMA were included, respectively), or if only morphologically abnormal embryos were included or could not be identified (Table S2). Five series were additionally reviewed manually and three were included38-40. The process of study review and selection is outlined in Figure 1.

The main characteristics of the 23 studies included in the meta-analysis are outlined in Table 1. Ten studies included only first-trimester pregnancy losses9, 20, 22, 23, 25, 30, 35, 37, 38, 40 and 13 included losses under 20 weeks19,21,24,26–29,31–34,36,39. In six studies, CMA was performed after a normal karyotype or QF-PCR result, in seven studies it was performed concurrently to karyotyping, and in the remaining 10 as a first-tier technique. Only for cases with normal karyotype was the incremental yield over karyotyping assessed. Seven studies included samples of recurrent pregnancy losses27,30,35–39. Quality assessment of the included studies using QUADAS-2 is shown in Figure 2. Funnel plots are provided as online supplementary material (Figures S1 and S2).
Study | Inclusion criteriato original study | Cases in original study (n) | Early pregnancy losses (n) | Included in our analysis (normal karyotype) (n) | Gestational age (weeks) | Criteria for CMA | CMA type |
---|---|---|---|---|---|---|---|
Schaeffer (2004)19 | POC samples | 41 | 41 | 25 | < 20 | In all concurrently to karyotype and FISH | GenoSensor Array 30 kit (Vysis/Abbot) |
Shimokawa (2006)20 | Miscarriage with normal karyotype | 20 | 19 | 19 | 5–12 | In all with normal karyotype | Customized with 2173 Fished BAC clones (GenePix) |
Borovik (2008)38 | POC samples | 80 | 54 | 17 | — | In all with normal karyotype | Customized with 3500 BAC/PAC DNA targets spaces 1-Mb intervals |
Warren (2009)21 | Fetal loss at 10–20 weeks | 35 | 30 | 30 | 10–20 | In normal or not performed karyotype | Spectral 2600 whole genome BAC array (PerkinElmer) and Agilent 244 K oligonucleotide array to confirm it |
Zhang (2009)22 | 1T miscarriage | 115 | 58 | 58 | < 13 | In normal karyotype or PCR-based genotyping | 244 K whole genome array-CGH (Agilent) |
Robberecht (2009)9 | Miscarriage | 103 | 103 | 67 | — | In all concurrently to karyotype | Customized with 3534 BAC clones (invitrogen) |
Deshpande (2010)23 | Pregnancy loss | 20 | 20 | 20 | — | In all concurrently to karyotype, and MLPA | Focus Cytochips (Blue genome) BAC array - 1-Mb resolution |
Rajcan-Separovic (2010)39 | RPL with at least one miscarriage | 27 | 27 | 27 | < 20 | In normal karyotype | Agilent 105 K oligonucleotids array-CGH |
Gao (2012)40 | Miscarriage | 100 | 100 | 51 | < 12 | In all concurrently to karyotype, FISH or QF-PCR | Platform from Agilent with 60-mer oligonucleotide probes |
Li (2013)24 | POC from couples with normal chromosomes | 81 | 81 | 81 | 1T (< 12) + 2T (< 19) | SNP array in all | HiScanSQ (Illumina) |
Wang (2014)28 | POC samples | 268 | 268 | 24 | > 7 | In all concurrently to karyotype | Oligo-SNP CMA, Affymetrix CytoScan HD |
Levy (2014)26 | Miscarriage < 20 weeks | 2400 | 1861 | 755 | < 20 | In all | Illumina CytoSNP-12 genotyping CMA |
Kooper (2014)25 | 1T, 2T and 3T intrauterine fetal death | 417 | 17 | 6 | 1T (< 12) | In normal karyotype or QF-PCR | Affymetrix GeneChip 250 k (NspI) SNP-array |
Kudesia (2014)27 | RPL | 20 | 16 | 16 | < 20 | Standalone in all | 2 x 400 K CGH-arrays (Agilent) |
Rosenfeld (2015)31 | Fetal demise | 535 | 142 | 73 | < 20 | In all concurrently with karyotype | BAC-based arrays, Signature Chip version 2-4 or oligonucleotide-based array, Signature ChipOS versions 1-4 (Signature Genomic Laboratories) alone or combined with SNP v3.1 /v4.0 (Roche/Agilent) |
Lin (2015)32 | 1T and 2T miscarriage | 155 | 133 | 49 | 1T (< 12) + 2T (< 19) | In all, in parallel with karyotyping | Affymetrix CytoScan 750 K arrays |
Romero (2015)29 | POC samples | 86 | 74 | 39 | < 20 | SNP-array standalone | CytoScan SNP-array (Affymetrix) |
Maslow (2015)30 | POC samples from patients with two or more pregnancy losses | 62 | 44 | 42 | < 13 | SNP-array standalone | Illumina CytoSNP-12 genotyping CMA |
Sahoo (2017)36 | RPL | 8118 | 7396 | 3479 | < 20 | In all | 3520 clones, BAC-clone-based array-CGH (CombiMatrix), 180334oligonuclotide probes oligonucleotide array (Agilent) or CytoSNP-850 K array (Illumina) |
Ozawa (2016)35 | POC samples | 15 | 13 | 6 | < 9 | CMA standalone | Three versions of BAC-based arrays, (Genome Disorder Array); 550 660 712 BACs |
Zhang (2016)33 | Abortion or stillbirth | 60 | 43 | 40 | < 20 | SNP-array standalone | HumanCytoScan 750 K array (Affymetrix) |
Wou (2016)34 | Miscarriage, stillbirth or postnatal death | 1071 | 535 | 319 | < 20 | In all | 8x60K ISCA v2.0 (AMADID 26370) oligonucleotide array (BlueGnome) |
Wang (2017)37 | 1T miscarriage | 551 | 535 | 265 | < 13 | In all | HumanCyto 12SNParray (Illumina) |
- Only first author of each study is given.
- 1T/2T/3T, first/second/third trimester; BAC, bacterial artificial chromosomes; CGH, comparative genomic hybridization; CMA, chromosonal microarray analysis; FISH, fluorescence in-situ hybridization; MLPA, multiplex ligation-dependent probe amplification; PAC, P1-derived artificial chromosome; PCR, polymerase chain reaction; POC, products of conception; QF-PCR, quantitative-fluorescent polymerase chain reaction; RPL, recurrent pregnancy losses; SNP, single nucleotide polymorphism.




The mean test success rate obtained by karyotyping was 68% (920/1352), while that by CMA was 92% (12 812/13 911) (Table S3). In the nine series reporting concurrent success rates, a 41% increase, accounting for 27 percentage points, was observed, from 68% (920/1352) (95% CI, 66–70%) by karyotyping to 95% (1202/1263) (95% CI, 94–96%) by CMA (Table 2). In the nine series with available results by conventional karyotyping, 3247 chromosomal anomalies were found in 6472 samples, resulting in a 49% (95% CI, 41–57%) pooled rate (Figure 3). Triploidy, accounting for 8% (95% CI, 5–10%) of the chromosomal anomalies, can be missed by standalone aCGH.
Study | Karyotype | CMA | Test success rate difference (%) | ||||
---|---|---|---|---|---|---|---|
Successful | Attempted | Test success rate (%) | Successful | Attempted | Test success rate (%) | ||
Borovik (2008)38 | 49 | 54 | 91 | 17 | 17 | 100 | 9 |
Warren (2009)21 | 11 | 41 | 27 | 30 | 35 | 86 | 59 |
Robberecht (2009)9 | 77 | 103 | 75 | 91 | 103 | 88 | 13 |
Zhang (2009)22 | 92 | 115 | 80 | 58 | 58 | 100 | 20 |
Gao (2012)40 | 86 | 100 | 86 | 86 | 86 | 100 | 14 |
Wang (2014)28 | 161 | 268 | 60 | 240 | 268 | 90 | 30 |
Kooper (2014)25 | 1 | 1 | 100 | 6 | 6 | 100 | 0 |
Rosenfeld (2015)31 | 310 | 515 | 60 | 522 | 535 | 98 | 38 |
Lin (2015)32 | 133 | 155 | 86 | 152 | 155 | 98 | 12 |
Total | 920 | 1352 | 68 | 1202 | 1263 | 95 | 27 |
- Only first author of each study is given.

A 2% (106/5507) (95% CI, 1–2%) pooled mean incremental yield of pathogenic CNVs by CMA over karyotyping was observed in the 23 series included in the meta-analysis, with a 0 to 17% range (Figure 4). After stratification of series according to inclusion of recurrent pregnancy loss, the mean incremental yield in the seven series with recurrent losses was 2% (66/3691) (95% CI, 1–2%) (Figure S3), while in the 18 series with no recurrent loss, it was 1% (40/1816) (95% CI, 1–2%) (Figure S4). Among 106 pathogenic CNVs overall, 48 were specified in 16 of the series (Table 3), their size ranged between 100 kb and 11 Mb, and the most frequently found were del22q11.21 (n = 4)26, 31, 32, 37, and dup10q26 (n = 3)19, 28, 37. The following CNVs were found in two cases each: del1p36.3328, 35, del7q11.2326, 37, dup11p15.526, del22p1326, and del3p2620, 37.

Study | Trimester | Karyotype | Pathogenic CNVs |
---|---|---|---|
Schaeffer (2004)19 | First | 46,XX | dup10qtel |
Shimokawa (2006)20 | First | 46,XX | del3p26,2-p26,3 |
Warren (2009)21 | First | Normal | dup5p15.33 |
Zhang (2009)22 | First | Normal | dup2q12 |
dup9q22.33 | |||
dup19p11,2 | |||
dup18p 11.31 | |||
del9p21,2 | |||
Robberecht (2009)9 | First | 46,XX | delXp22.31 |
Gao (2012)40 | Failed | del7p21.3-p22.3(11.72 Mb) | |
Li (2013)24 | First | No data | 46,XX,del (3)(p21,31- > p21,1) |
No data | 46,XX,del (22)(p13,1- > p13,2) | ||
No data | 46,XX,del (3)(p21,31- > p23,31) | ||
No data | 46,XX,del (17)(q22- > q23,3) | ||
No data | 46,XX,del (20)(p12.1- > p12.2) | ||
No data | 46,XX,dup, (8)(p23,1- > 23.2) | ||
Wang (2014)28 | — | Normal | arr [hg19]1p36.33p36.21(849466-15 970 926) × 1, 10q26.2q26.3(129968527-135 427 143) × 3 |
Levy (2014)26 | First | Normal | del1q21.1-q21.2 (144397794-148 445 751) |
Normal | del7q11.23(72000000-75 000 000) | ||
Normal | del14q32.2-q32.33(99039627-106 368 585) | ||
Normal | del15q11.2(18450000-21 240 000) | ||
Normal | del16p11.2(29830000-31 430 000) | ||
Normal | del18p11.32-p11.23(0-7 956 303) | ||
Normal | del22q11.21(16980-20 130 000) | ||
Normal | del22q13,2q13.33(42520000-49 510 000) | ||
Normal | del22q13,2q13.33(40020000-49 510 000)) | ||
Normal | dup8q24.23q24.3(139000000-146 200 000) | ||
Normal | dup11p15.5-p15.4(0-6 638 049) | ||
Normal | dup11p15.5-p15.4(193788-9 388 462) | ||
Kooper (2014)25 | First | No data | arr3p24.2p24.3(17269256-25 630 783) × 1 |
Rosenfeld (2015)31 | First | 46,XY | arr[hg19]Xp21.1(31870020-31 949 549) × 0mat,17q25.1q25.3(73080340,-81 043 894) × 3dn |
Second | 46,XY | arr[hg19]22q11.21(18919469-21 460 658) × 1 | |
Lin (2015)32 | First + second | Normal | arr del22q11.21 (2,88 Mb) |
Ozawa (2016)35 | First | No data | 1p36.33p36.32 |
Zhang (2016)33 | First | 46,XY | 15q21.1, gain |
Wou (2016)34 | First | Normal | arr 12q21.2q21.31(79876110-84 526 577) × 1 dn |
Wang (2017)37 | First | Normal | arr[hg19]7q36.1q36.3(152050571-159 119 486) × 17q35(144408433-147 566 163) × 3 |
arr[hg19]3p26.3p26.1(66894-5 967 798) × 1 | |||
arr[hg19]7p22.3p22.1(46239-5 820 350) × 1 | |||
arr[hg19]10q26.13q26.3(125498896-135 063 876) × 3 | |||
arr[hg19]2q37.2q37.3(236972635-243 029 573) × 1 | |||
arr[hg19]19p13.3(267039-1 771 582) × 1 | |||
arr[hg19]22q11.21(18938367-21 462 353) × 1 | |||
arr[hg19]16p13.12p13.11(14760734-16 303 388) × 1 | |||
arr[hg19]Xp22.31(6516735-8 091 350) × 3 | |||
arr[hg19]7q11.23(72350815-74 282 048) × 1 | |||
arr[hg19]3q29(195738406-197 421 203) × 1 | |||
arr[hg19]16p13.11(15052746-16 633 361) × 1 |
- Only first author of each study is given.
Twelve studies provided information on VOUS21, 25-29, 31, 33, 34, 36, 37, 39, demonstrating a 4% (218/5072) (95% CI, 3–6%) VOUS yield by CMA (Figure 5). Among 218 VOUS overall, 69 were specified (Table S4), and their size ranged between 0.4 kb and 3.65 Mb, smaller than that of pathogenic CNVs. In the five studies in which a SNP-array was applied, uniparental disomy was found in 1.1% (58/5086) (95% CI, 0.8–1.4%) of the samples26, 28, 32, 36, 37.

DISCUSSION
This systematic review and meta-analysis supports the use of CMA as the test of choice for investigation of genetic causes of pregnancy loss up to 20 weeks. CMA was shown to be superior to conventional cytogenetics, mainly due to a significant increase, by 41%, in the test success rate (from 68% to 95%), in addition to at least a 2% (95% CI, 1–2%) incremental diagnostic yield over conventional karyotyping.
Diagnosis of chromosomal anomalies in pregnancy loss provides important information for genetic counseling on recurrence risk, given that it varies as to whether a familial chromosomal rearrangement or an autosomal trisomy is identified. It is well known that 50–70% of non-recurrent pregnancy losses are due to aneuploidy4, 5, but assessing the chromosomes after a pregnancy loss has been technically challenging. Cytogenetic analysis not only requires culture of live cells, but carries a risk of erroneous results due to maternal cell contamination8-12. CMA also carries a risk of maternal cell contamination that can be substantially ruled out if a previous QF-PCR is performed or a SNP-array is used. Nevertheless, although SNP-arrays can detect confidently deletions in such samples, caution should be taken since a gain in a sample with maternal cell contamination might be missed. Interestingly, the addition of a previous QF-PCR if aCGH, instead of SNP-array, is applied may be useful also to detect triploidy and other polyploidies. Our group routinely offers genetic testing to women suffering a pregnancy loss, irrespective of number of previous losses7. Knowing the cause of the miscarriage and being aware of the recurrence risk can help women to cope with the loss. In addition, in recurrent losses, the fact that karyotypes of previous losses are already disclosed, simplifies the work-up if a chromosomal anomaly was revealed in all previous losses.
Regarding the incremental yield of CMA over karyotyping, our systematic review found 23 series that fulfilled the search criteria, with two large series including 755 and 3479 cases, sharing a 2% incremental yield26, 36. The remaining series were smaller, including from 6 to 319 cases, and reported a 0–17% incremental yield9,19-25,27-35,37–40. Although initially surprising, it can be argued that the low incremental yield of CMA observed in early pregnancy loss can be explained by the fact that early losses are expected to be due to large dosage changes. Microdeletions and microduplications are expected to be more strongly related to intellectual disability and neurocognitive impairment in surviving individuals.
The previous meta-analysis regarding the added value of CMA over conventional karyotyping by Dhillon et al. in 201413, revealed a 13% (95% CI, 8–21%) increase in chromosome anomalies detected by CMA compared with karyotyping. This meta-analysis included nine series with 314 cases overall, reported before 2013. All but one series have been included in our meta-analysis. The excluded study contained only samples from embryos with abnormal morphology. The large discrepancy between the two meta-analyses (13% vs 2%) may have several explanations. Firstly, the 13% incremental yield in Dhillon's study included both pathogenic CNVs and VOUS. Secondly, their meta-analysis included only nine small studies (ranging from 10 to 86 cases) with a total of 314 cases, while ours included 23 studies with a total of 5507 cases.
The most frequently encountered pathogenic imbalance revealed by our meta-analysis was the 22q11.2 microdeletion syndrome, also named DiGeorge/velocardiofacial syndrome. This syndrome is the second most common cause of developmental delay and major congenital heart disease after Down syndrome, accounting for approximately 2.4% of individuals with developmental disabilities, and approximately 10–15% of patients with tetralogy of Fallot41. The observed prevalence of the 22q11.2 microdeletion was 0.22% (4/1813), concordant with previously reported data on early pregnancy losses26, 31, 32, 37, and higher than the one reported in the general population (0.013% of live births)42. Other common microdeletion syndromes were 1p36 and 7q11.23. The 1p36 microdeletion syndrome is one of the most common deletion syndromes, characterized by craniofacial dysmorphism, heart defects and renal and skeletal anomalies43. Another commonly reported microdeletion is 7q11.23, which causes Williams–Beuren syndrome44. However, it is unclear whether these microdeletions were the cause of miscarriage, since the 95% CI of the observed 2% incremental yield, from 1% to 3%, overlaps the prevalence of pathogenic CNVs expected in the general population. Unfortunately, no information is reported on the proportion of inherited changes.
This is the second systematic review and meta-analysis on the incremental yield of CMA over karyotyping in early pregnancy loss. The main strength of our study is its size when compared with the first meta-analysis (23 vs 9 studies, 5507 vs 314 cases). The main limitation is the heterogeneity of included studies with regards to the clinical study design (CMA performed either concurrently with karyotyping, after a normal karyotype result, or after normal FISH or QF-PCR results), CMA platform used (bacterial artificial chromosomes, oligonucleotides in aCGH, or SNP-arrays), year of publication (between 2004 and 2017), indications for analysis, and inclusion of early second-trimester miscarriages, given that the upper limit of the gestational age period ranged from 12 to 20 weeks.
In conclusion, this systematic review and meta-analysis revealed that CMA provides informative results in a significantly higher proportion of early pregnancy losses than does conventional karyotyping, because it delivers an extra 41% of informative results (95% vs 68%), with at least 2% incremental diagnostic yield. The use of SNP-array instead of aCGH may detect maternal cell contamination, triploidy and uniparental disomy (otherwise a previous QF-PCR may solve the first two issues). Taking into account its higher success rate and incremental diagnostic yield, CMA testing should be considered as the first-tier test for genetic investigation of pregnancy loss.
ACKNOWLEDGMENTS
The project received a research grant from the Carlos III Institute of Health, Ministry of Economy and Competitiveness (Spain), under the Health Strategy Action 2013–2016, with reference PI14/00588, cofunded with European Union ERDF (European Regional Development Fund) funds.