Approximately 50% of spontaneous miscarriages are associated with chromosome abnormalities. Identification of these karyotypic abnormalities helps to estimate recurrence risks in future pregnancies. Chromosomal microarray analysis (CMA) is transforming clinical cytogenetic practice with its ability to examine the human genome at increasingly high resolution.
The aim of this study was to determine whether CMA testing on the products of conception following miscarriage provides better diagnostic information compared with conventional karyotyping.
MEDLINE (from 1996 to December 2012), EMBASE (from 1974 to December 2012), and CINAHL (from 1996 to December 2012) databases were searched electronically.
Studies were selected if CMA was used on products of conception following miscarriage, alongside conventional karyotyping.
Data collection and analysis
Nine papers were included in the systematic review and meta-analysis. All statistical analyses were performed using stata 11.0 (Stata Corp., College Station, TX, USA).
There was agreement between CMA and karyotyping in 86.0% of cases (95% CI 77.0–96.0%). CMA detected 13% (95% CI 8.0–21.0) additional chromosome abnormalities over conventional full karyotyping. In addition, traditional, full karyotyping detected 3% (95% CI 1.0–10.0%) additional abnormalities over CMA. The incidence of a variant of unknown significance (VOUS) being detected was 2% (95% CI 1.0–10.0%).
Compared with karyotyping, there appears to be an increased detection rate of chromosomal abnormalities when CMA is used to analyse the products of conception; however, some of these abnormalities are VOUS, and this information should be provided when counselling women following miscarriage and when taking consent for the analysis of miscarriage products by CMA.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Approximately 10–15% of all clinically recognised pregnancies end in miscarriage, most of which occur in the first trimester. Of these first-trimester miscarriages, ~50% are caused by chromosome abnormalities. The majority of these (86%) are numerical abnormalities, including trisomies, monosomies, and polyploidies (triploidy or tetraploidy). Structural chromosomal abnormalities (large duplication or deletions visible by full conventional karyotyping) account for a further 6%, and other abnormalities such as single gene mutations and mosaicism account for 8%. The identification of the cause of spontaneous miscarriage helps to estimate recurrence risks in future pregnancies, and when an anomaly is found this is useful information to provide when counselling parents.
Chromosome studies on the products of conception are indicated after the third and subsequent consecutive miscarriages. After analysis, in ~2–5% of these couples one partner will be identified as carrying a balanced structural chromosomal anomaly, most commonly a balanced reciprocal or Robertsonian translocation. Although carriers of balanced translocations are phenotypically normal, their pregnancies are at greater risk of miscarriage if the resulting pregnancy has an unbalanced chromosomal arrangement.
Chromosomal microarray analysis (CMA) has advantages over conventional chromosome analysis (i.e. the karyotyping of the metaphase spreads after tissue culture), including improved resolution and potentially higher detection rates of chromosomal abnormalities. The benefits and challenges of CMA for studying miscarriage are outlined in Table 1. A high-resolution oligonucleotide array is capable of detecting changes to a resolution of 1 kb (smaller than the average gene).
Table 1. Benefits and challenges of chromosomal microarray for studying miscarriages
Detects most types of large chromosomal imbalances
Neither array type (SNP or CGH) can detect balanced rearrangements: the CGH array is more limited for the detection of ploidy change than the SNP array
Arrays are quick and DNA based; DNA from paraffin-embedded tissue can be used
DNA tends to be poor quality and degraded
Arrays are DNA based, so the effects of tissue culture failure and artefacts of maternal contamination are minimised
If tissue culture is omitted, there are no chromosomes for FISH confirmation; CNV confirmation has to be DNA based
Microarrays detect potentially pathogenic CNVs that cause or contribute to miscarriage
More parental investigations are needed to interpret CNVs. Therefore, more couples will have to be investigated to determine whether they carry a miscarriage CNV
Improves our understanding of genetic and biological factors implicated in early human development
More extensive genetic counselling is necessary for uncertain findings
Chromosomal microarray analysis (CMA) does have potential limitations, however: the most significant disadvantage is the identification of novel, previously unreported, variants of unknown significance (VOUS) and the difficulties this may cause for clinical management. In order to facilitate a more accurate assessment and interpretation of VOUS there has been a major effort to catalogue and collate genomic and clinical information.
As CMA technology will inevitably become increasingly important as a diagnostic tool for detecting chromosome abnormalities in the products of conception, we sought to systematically review the literature and describe the evidence for debating whether CMA offers greater diagnostic information over conventional karyotyping.
Our systematic review followed a prospective protocol developed using widely recommended and comprehensive methodology.[4-6]
The search focused on studies using CMA compared with conventional karyotyping on the products of conception following miscarriage. A search strategy was developed based on existing advice for prevalence searches.[5, 6] MEDLINE (from 1996 to December 2012), EMBASE (from 1974 to December 2012), and CINAHL (from 1996 to December 2012) databases were searched electronically, and Web of Science was used to search for grey literature (literature not conventionally published and not widely available). The search of MEDLINE and EMBASE captured citations containing the relevant medical subject headings (MeSH) and word variants for ‘microarray’ and ‘miscarriage’. The following terms were used to describe microarrays: array, comparative, genomic, hybridization, and microarray. Similarly, abortion spontaneous, abortion habitual, recurrent miscarriage and pregnancy loss were used to capture ‘miscarriage’.
The bibliographies of relevant articles were manually searched to identify papers not captured by the electronic searches. Authors were contacted for completeness of the search. There were no language restrictions in the search or selection of papers.
Eligibility criteria for selecting studies
Studies were selected in a two-stage process. Initially, all abstracts or titles in the electronic searches were scrutinised by two reviewers (R.D. and S.H.), and full manuscripts of potentially eligible citations were obtained. Differences were resolved by discussion with a third reviewer (R.K.M.). Studies were selected if CMA had been used to diagnose abnormalities in the products of conception following miscarriage and had also included standard karyotyping (T or G banding). We included papers in which the CMA was performed only on normal karyotypes as well the studies that performed the array on both normal and abnormal karyotypes. Papers were excluded if they used the CGH technique and not CMA. We excluded papers where multiple diagnostic techniques were employed producing data from which CMA-specific data could not be extracted. Only papers that allowed the generation of a 2 × 2 table (comparing karyotyping with CMA) were included. A summary of the selection process is displayed in Figure 1.
Data were extracted by two reviewers (R.D. and S.H.). For each of the outcomes, data were extracted into tables. Data were extracted on study characteristics and data quality. Data were used to construct 2 × 2 tables of diagnostic test accuracy comparing normal and abnormal karyotype results against normal and abnormal CMA results (Table S1). Case studies of fewer than five cases were excluded from the meta-analysis.
All articles meeting the selection criteria were assessed using items from validated tools. Quality assessment tool for diagnostic accuracy studies (QUADAS), a methodology checklist tool for studies of diagnostic test accuracy, was used to assess study quality (Figure 2). Expert opinion from a senior cytogeneticist (D.M.) at the West Midlands Regional Genetics Department was sought to determine the validity of the array used.
The analysis was performed by grouping array results that were pathogenic and VOUS, and therefore potentially clinically relevant, together as positive results. Analysis was performed for samples undergoing both karyotyping and CMA. There were no benign copy number variants (CNVs) reported. The possibility of separating the data for cases where CMA and karyotyping were both abnormal, but where one provided additional information over the other, was considered, but as the numbers were so few it was decided to leave the data combined and comment on this in the results. Using 2 × 2 tables, we computed and pooled the percentage agreement between the two technologies (with 95% CIs) for the studies overall. Then calculated the percentage of extra cases identified by the array in those with a normal karyotype, and vice versa, with a 95% CI. We also calculated the estimated VOUS rate with a 95% CI; Table S2 shows the number of VOUS cases found in each paper. The heterogeneity in rates was examined graphically and statistically. For graphical assessment, forest plots of point estimates of rates and their 95% CIs were used. Funnel plot asymmetry was assessed using Egger's test.
For the meta-analysis, log rates were pooled, weighting each study by the inverse of its variance, and the summary estimates were exponentiated. A random-effects model was used where there was statistically significant heterogeneity. All statistical analyses were performed using stata 11.0 (Stata Corp., College Station, TX, USA).
Nine primary articles were identified as meeting the selection criteria.[8-16] Of the original 239 articles found using our search terms, only 33 met the selection criteria. Of these 33, we obtained the full articles for 32, and we contacted the authors of the abstract written by McNamee et al., but the full paper was not provided as the data are yet to be published. From these 32 articles, 20 were removed as there were no original data available to extract and 12 were found to be suitable for inclusion. However, on closer evaluation, we removed a further three publications as extractable information for 2 × 2 tables was not possible.[18-20] This led to the inclusion of nine papers for the final systematic review.
Table 2 summarises the characteristics of the publications used, including the study design (prospective or retrospective), the array type, the sample type, the indication for the array, and the sample size.
Table 2. Summary of included study characteristics
Oligonucleotide-based aCGH with a 244-K chip (Agilent Technologies)
Products of conception
Normal karyotype detected by G-banding or PCR-based genotyping
Initial total of 115 miscarriages: 92 analysed by G-banding, of these 37 were normal. Remaining 23 cases where G-banding not possible, PCR-based genotyping performed, with two cases removed for triploidy
Total cases analysed by CMA and G-banding: n = 37 cases
Customised, with 3500 BAC/PAC DNA targets spaced at 1-Mb intervals
Products of conception
First-trimester spontaneous miscarriage with normal karyotype
Initially 49 cases: 22 abnormal karyotypes removed as array performed on normal karyotypes only; ten removed as no suitable material; three cases removed as ambiguous information in table and discussion (cases 17, 24, and 51)
Initially 100 cases: karyotyping failed in 14, with remaining cases analysed
n = 86 cases
The overall agreement between karyotype and CMA results was 86.0% (95% CI 77.0–96.0). This was using a total of 314 samples and included all nine articles.[8-16] Figure 3 shows the agreement between CMA and karyotyping. These results were homogeneous throughout the studies (Cochrane Q, I2 = 0.2%).
The CMA detected 13% (95% CI 8.0–21.0) additional chromosome abnormalities over karyotyping (Figure 4). Four studies performed an array on normal karyotypes only,[8-10, 14] whereas the remaining five performed CMA on both normal and abnormal karyotypes.[9, 11, 12, 15, 16] For the purpose of this systematic review these papers were combined using all 317 cases in the meta-analysis. Results that were interpreted as VOUS were grouped with results that were pathogenic. No cases of benign chromosomal abnormalities were reported. These data were heterogeneous (Cochrane Q, I2 = 92.4%).
Full conventional karyotyping detected 3% (95% CI 1.0–10.0) additional abnormalities over CMA (Figure 5). For the meta-analysis only the papers in which an array was performed on both normal and abnormal karyotypes were included, giving a total of 233 cases.[9, 11, 12, 15, 16] These data were heterogeneous (Cochrane Q, I2 = 98.5%).
These percentages may be lower than the actual detection rate of information provided by CMA, compared with karyotyping, as in four studies where positive results were found for both CMA and karyotyping, either the CMA or the karyotyping provided extra information in addition to this positive result (Table 3). In the paper by Robberecht et al., of the 15 cases where both CMA and karyotyping gave a positive result, in one case (Sample 3) the karyotyping found a marker chromosome not detected by CMA. However, CMA provided extra information over karyotype in six of these positive 15 cases. In the Schaeffer paper, three out of an agreed positive 16 cases array provided extra information over karyotyping, this includes a Trisomy 20, a deletion at 9p21 and a duplication at 15q. In the paper by Deshpande et al., the CMA and karyotyping agreed positively in five cases, and CMA provided extra information in just one case detecting mosaicism. In Gao et al. paper,16 44 cases were positive by both tests, but in one case a trisomy 22 detected by karyotype was instead found to be a mosaicism for trisomy 22 and trisomy 9 by detailed microarray analysis.
Table 3. Extra information provided by CMA or full karyotyping in cases (from the reviewed literature) where both tests gave positive findings
Five studies had VOUS,[8-11, 15] and pooling these figures showed the chance of having a VOUS was 2% (95% CI 1.0–10.0%; Figure 6). This summary estimate was heterogeneous (Cochrane Q, I2 = 99.7%).
Assessing for funnel asymmetry using Egger's test showed no bias for both the results for CMA over karyotyping, and vice versa: P = 0.3863 and P = 0.6644, respectively. There was no funnel asymmetry seen in the results for agreement or VOUS rate: P = 0.0989 and P = 0.7329, respectively.
As expected, we found good overall agreement between CMA results and conventional karyotyping from the literature obtained.
The CMA detected significantly more chromosomal abnormalities than conventional karyotyping; however, a proportion of these were VOUS. There were three studies (Robberecht et al. Deshpande et al. and Gao et al.) where karyotyping was positive with a normal array. For Robberecht and Deshpande this occurred in three samples, and all were balanced translocations; however, Gao et al. detected three triploids diagnosed by karyotype where the CMA was normal, and two tetraploid cases that proved to be culture artefacts. The inability of CMA to detect truly balanced rearrangements is well recognised; however, the clinical significance of such a result is questionable in terms of its causality for miscarriage.
In the cases where both full conventional karyotyping and CMA gave positive results, but one gave greater information, advice was sought from a consultant senior molecular cytogeneticist (D.W.) as to whether these findings would alter the counselling of the couple (Table 3). The conclusion was that in the three cases where the chromosome result in the products of conception could have resulted from a parental balanced chromosome rearrangement, this would have affected the counselling. These were Schaeffer et al. cases 29 and 38, and Gao et al. case 74 (Table 3). Here, the additional findings of del9p21 and dup15q, respectively, may be the product of a balanced translocation carried by one of the parents and parental chromosomes would be required. The finding of mosaicism for trisomy 22 and trisomy 9 in the Gao et al. paper is clinically significant, and thus would affect counselling.
Karyotyping did not detect any pathogenic chromosomal imbalances when CMA was reported as normal; however, CMA will not detect polypoidies, and therefore there may have been some selection bias in the reporting of cases. Overall this supports the idea that CMA alone will not miss many significant results, and that areas in which the technology is known to be weak (low-level mosacism and a lower sensitivity for triploidy) are uncommon in a typical representative population.
Strengths and weaknesses
The VOUS occur in approximately 2% of cases; however, the confidence intervals surrounding this were large, and more data are needed to establish the true VOUS rate. The interpretation of the VOUS finding is a limitation to this study. For the purposes of the analysis, cases of VOUS were taken as pathogenic, unless it could be proven to be benign. Therefore, the significance of the microarray detecting clinically and pathogenically significant abnormalities over karyotyping has to be interpreted with a degree of caution.
One of the strengths of this study was the thorough methodological approach used. It met the quality criteria laid down in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement. The analysis contained a relatively small sample size, with 313 cases for overall analysis; however, as a large majority of studies published single case studies, this represents a reasonable size of cohort. The papers were heterogeneous with both prospective and retrospective methodology. There was wide variation between the papers: different CMA platforms (customised versus commercially available); year of publication, which is important given the rapidly progressing field; and clinician bias influencing case selection. The papers also had slightly different indications for referral. This would account for the large confidence intervals in our analysis. Using the QUADAS checklist to diagnose the accuracy of the studies, we found that all of the papers met the majority of the criteria required, and were therefore of sound quality.
One of the limitations to this review was the inclusion of studies where only known normal karyotypes were chosen. This may have biased the results, as all the aberrations that CMA would have detected were over normal karyotypes, and thus not necessarily from a representative population. Therefore, this makes it difficult to apply the findings in the case of abnormal karyotypes.
The CMA technology continues to develop, with increasing resolution. We are unable to account for potential clinician bias in circumstances where it was felt that a case would yield an abnormal result from CMA, and was therefore included in the study. It is recognised that array technology is a developing scientific field with ever expanding literature over time. The data from this review represents a fair critical appraisal of the literature to date, and provides an accurate current summary.
Interpretation of findings
Further prospective research is needed in this area using a large cohort, with a representative, prospective population undergoing both a recognised reproducible array and karyotyping. In addition to this, determining whether the CNVs found in the products of conception are pathogenic or benign, based on a large cohort of miscarriage cases, is also required. CMA analysis of miscarriages is potentially advantageous compared with traditional cytogenetics. Cell culture is usually not required, which potentially avoids culture failure, culture contamination, overgrowth with maternal cells (contamination) or selection against the chromosomal abnormal cells of mosaic fetuses. CMA has a higher resolution than conventional karyotyping, thus allowing for the detection of small deletions and duplications. The resolution of recently developed arrays has been steadily increasing, and is only limited by the size and distance between different probes. The cost-effectiveness depends on where the array is placed in the cytogenetic testing process: CMA is often used in combination with other cytogenetic tests, and as an adjunct test, rather than a stand-alone diagnostic test.
A recent literature review by Rajcan-Separovic concluded that despite the high potential to improve genetic analysis, the reported use of CMA for studying miscarriages is still limited in comparison with its widespread use to test patients with developmental delay/congenital anomalies, pre-implantation embryos, and continuing pregnancies.
This systematic review provides accurate evidence of the relative advantage of using array testing over karyotyping in diagnosing chromosomal abnormalities in the products of conception. The additional detection of CNVs by array technology is a combination of known pathological findings and VOUS. This information may be useful in counselling couples following miscarriage. With information being continuously added to databases, e.g. the database for genomic variation (DGV), areas of uncertainty, such as the prevalence of CNVs between differing ethnic populations, will undoubtedly diminish.
There appears to be an increased detection rate of chromosomal variants with CMA over conventional karyotyping in the analysis of products of conception. The large confidence intervals found in this analysis demonstrate that further work is required before we can answer the absolute detection rate.
Nonetheless, CMA analysis will expand the possibilities to support clinical management of the couples experiencing miscarriages, the identification of recurrent pregnancy loss (RPL), associated CNVs and genes, and to develop new knowledge of the factors and causes of early failure of human development.
Disclosure of interests
Contribution to authorship
R.D. and S.H. reviewed the literature. R.K.M. was the third reviewer. R.D. and S.H. performed statistical analysis, and R.K.M. assisted with the statistical analysis. R.D. wrote the article. D.M. reviewed cases of VOUS. D.W. advised on impact on counselling. A.C. and M.K. contributed to and reviewed the final article.
Details of ethics approval
No formal ethical approval required.
RD is funded by the EME TABLET Trial. SH is funded by the Children's Charity SPARKS. RKM is funded by NIHR clinical lecturer fellowship.
Commentary on ‘Additional information from chromosomal microarray analysis (CMA). over conventional karyotyping when diagnosing chromosomal abnormalities in miscarriage: a systematic review and meta-analysis'
The meta-analysis described by Dhillon et al. has shown that chromosomal microarrays (CMAs) detected 13% more genetic abnormalities than conventional karyotyping of the abortus. CMA easily identifies aneuploidy, unbalanced translocations, deletions, and duplications; however, balanced translocations, haploidy, or polyploidy cannot be detected. Indeed, as described by Dhillon et al., traditional karyotyping detected 3% additional abnormalities that were not detected by CMA. Therefore, for a complete picture, it may be necessary to perform a standard karyotype in addition to CMA.
Many genetically abnormal embryos cease developing, either at pre-implantation stages, presenting as recurrent implantation failure (RIF), or at later stages, presenting as missed miscarriage. CMA may be potentially useful in pregestational screening (PGS), whether performed for RIF or recurrent miscarriage. Until now PGS has not proved itself. Some embryos are injured by the embryo biopsy, and hence the pregnancy rates are lower. Additionally, prior to CMA, only a few chromosomes were screened in order to exclude the common chromosome aberrations (trisomies 13, 16, 18, and 21, monosomy X, and triploidy). Aberrations involving other chromosomes were missed. Pregestational diagnosis (PGD) has been compared with natural conception in women with recurrent miscarriage and with a parental structural aberration (Musters et al. Fertil Steril 2011;95:2153–2157). As expected, PGD lowered the incidence of miscarriage; however, because of the reduced pregnancy rate after PGD, there were fewer live births. (After natural conception, 42% of pregnancies result in live births, with a miscarriage rate of 28%. After PGD, 35% of pregnancies result in live births, with a miscarriage rate of 9%.) PGS diagnoses a single cell only. Approximately 75% of preimplantation embryos may be mosaic (Van Echten-Arends et al. Reproduction Update 2011;17:620–627). Li et al. (Fertil Steril 2005;84:1395–1400) found that 40% of the embryos diagnosed as aneuploid on day 3 had a euploid inner cell mass on day 6. Consequently, normal embryos may be discarded.
If PGS is to have its maximal value, the entire genome should be tested to exclude the 13% extra anomalies that Dhillon et al. described; however, CMA is expensive, and still has major problems. Whole genome amplification of single cells suffers from amplification bias and allele drop-out (ADO). ADO limits the detection of copy number variants (CNVs), and may generate many sequencing errors. Approximately 12% of the human genome exhibits copy number variation (Redon et al. Nature 2006;444:444–445). Athough some variations are normal and some are abnormal, there are a number of variants of unknown significance (VOUS). These complicate diagnosis, and require parental array analysis in order to determine which mutations arise naturally. In VOUS variations, the embryos are not currently selected for replacement at in vitro fertilisation, thereby limiting the number of embryos available for transfer. Dhillon et al. considered VOUS variations to be pathological in their metaanalysis. CNV databases are being developed, to classify CNVs that are presently VOUS as normal or pathogenic.
As a result of the above drawbacks, it is unclear whether PGS with CMA is warranted in RIF or recurrent miscarriage. At present, the European Society of Human Reproduction and Embryology (ESHRE) is supporting a multicentre randomised trial to determine whether PGS using CNV increases the delivery rates in women of advanced maternal age. The results are eagerly awaited, and should also have implications for RIF and recurrent miscarriage.
Declaration of interests
The author has no conflict of interests in this paper.
Department of Obstetrics & Gynecology, Sheba Medical Center, Tel Hashomer & University of Tel Aviv, Israel