Karyotype versus genomic hybridization for the prenatal diagnosis of chromosomal abnormalities: a metaanalysis




Materials and Methods


This study was conducted according to the recommendations of the Cochrane Collaboration and is reported following the PRISMA Statement. The protocol was registered in the international prospective register of systematic reviews (PROSPERO): CRD42014007627.


We designed a search strategy for studies published in MEDLINE via PubMed, CENTRAL, Cochrane Register of Diagnostic Test Accuracy Studies, and EMBASE. The search strategy was specific for each database and included a combination of the medical subject headings and free-text terms for “comparative genomic hybridization” and “karyotype.” No language or publication status restrictions were imposed. We included articles from Jan. 1, 1980, through March 31, 2014. The full search strategies are listed in the Appendix .


Other electronic sources were used to find additional studies, such as conference abstracts, Google Scholar, DARE, and PROSPERO. We looked for additional studies in the reference lists of selected articles and contacted authors about their knowledge of published or unpublished articles. The results of searches were cross-checked to eliminate duplicates.


Eligibility criteria


Studies


We included cross-sectional, case-control, and cohort studies conducted from Jan. 1, 1980, through March 31, 2014. No language restrictions were imposed. Studies were required to report at least sensitivity and specificity or data to calculate these parameters.


Participants


Pregnant women who underwent chorionic villus sampling, amniocentesis, or cordocentesis to perform CGH and karyotyping.


There were no preferences with respect to any other demographic characteristics of the participants.


Comparisons


We intended to perform the following comparisons:




  • Karyotype (reference standard) vs CGH (index test).



  • Karyotype plus CGH vs karyotype.



  • Karyotype plus CGH vs CGH.



However, at the time of analysis, we determined that CGH diagnosed abnormalities that the karyotype did not. Therefore, we decided to create a reference standard according to the literature (karyotype + CGH).


Outcomes


Outcomes were sensitivity, specificity, and likelihood ratios for numeric and structural chromosomal abnormalities.


Exclusions


We excluded studies using karyotype or CGH independently, and those in which data were unavailable to obtain sensitivity and specificity.


Study selection


Two investigators (H.A.G-P., J.A-P.) independently and blindly screened the titles and abstracts to determine the potential usefulness of the articles. Two assessors (H.A.G-P., J.A-P.) applied eligibility criteria to the full-text articles during the final selection. When discrepancies occurred, a final decision was reached by consensus. If the 2 assessors could not agree, a third reviewer (W.S.) made the final decision.


Data collection process


Relevant data were collected using a standardized data extraction sheet, which contained the following: study design, methods, participants, index test, standard of reference, and final outcome details. Reviewers confirmed all data entries and checked the entries at least twice for completeness and accuracy. If some information was missing, we contacted the authors to obtain the complete data.


Risk of bias in and across individual studies


The risk of bias was assessed independently by at least 2 researchers (H.A.G-P., J.A-P.) using the QUADAS2 tool, which evaluates items related to the patient selection, the index and reference tests, the flow and timing, and the concerns about their applicability. We solved disagreements by consensus. The risk of bias table (within and across studies) was edited using Review Manager Software version 5.2 (RevMan; Cochrane Collaboration, Oxford, England) to illustrate the judgments for each study.


Summary measures


Analyses were performed in RevMan 5.2, OpenMeta[Analyst] ( http://www.cebm.brown.edu/open _meta), and Stata 10 (StataCorp, College Station, TX) as needed. The sensitivity, specificity, likelihood ratios, and diagnostic odds ratios were measured for comparisons with 95% confidence intervals (CIs). We performed fixed effects or random effect metaanalysis according to the heterogeneity or homogeneity among the studies. We also performed forest plots and summary receiver operating characteristic for comparisons.


Heterogeneity between trials was assessed through the I 2 statistic. A value ≥50% can represent heterogeneity according to Higgins and Green (2011). We also intended to analyze heterogeneity according to the following: reference standard, clinical spectrum, type of method used, and age of the patient.


Additional analyses


We intended to perform the following subgroup analysis: low- and high-risk pregnancies, history of chromosomal abnormalities, parents with chromosomal abnormalities, maternal age >37 years, biochemical screening plus maternal ultrasonography, and abnormalities detected on ultrasonography. However, the studies lacked sufficient data to perform these types of analyses.


Sensitivity analysis


We undertook a sensitivity analysis based on unknown significance variables considered important for analysis and results.


Publication bias


Publication bias was not assessed due to the number of studies found (<10 studies) according to Higgins and Green.




Results


Study selection


In all, 137 articles were found with the described search strategies; after exclusions, 6 and 5 studies were included in qualitative (general description of the data) and metaanalysis, respectively (Armengol et al [2012], Fiorentino et al [2011], Maya et al [2010], Wapner et al [2012], and Van den Veyver et al [2009]) ( Figure 1 ).




Figure 1


Flowchart

CGH , comparative genomic hybridization.

Saldarriaga. Karyotype vs genomic hybridization for the prenatal diagnosis of chromosomal abnormalities. Am J Obstet Gynecol 2015 .


Characteristics of included studies


In all, 9974 pregnant patients were identified in the studies, with a median of 971 (interquartile range, 269–4282) patients per study. With respect to the array platform used in the various studies, the majority of studies used 1 of 2 arrays for CGH: Lee et al (2012) initially (until 2010) performed arrays based on bacterial artificial chromosomes (BAC) with 1-Mb resolution and later used oligonucleotide arrays. Van den Veyver et al (2009) analyzed 190 samples using BAC, and the rest were analyzed with oligonucleotides. Maya et al (2010) followed a similar protocol, although they later used whole-genome coverage BAC. In contrast, Fiorentino et al (2011) analyzed samples by means of a single platform. One study did not describe the type of array used ( Table 1 ).



Table 1

Characteristics of included studies





















































Study Country Array type Sample type Array indication Sample size, n
Van den Veyver et al (2009), prospective (cross-sectional) United States BAC chromosomal microarray V5 or 6; V6 of BCM oligonucleotide chromosomal microarray AF 254, CVS 53 Advanced maternal age (33.5%), abnormal ultrasound finding (22.9%), family history of genetic disease (23.7%), abnormal fetal karyotype (7.6%), parental anxiety (9%), altered serum screening (2.5%), others (0.9%) 309
Maya et al (2010), retrospective (cross-sectional) Israel BAC using SignatureChip whole genome or oligonucleotide microarrays AF 243, CVS 16 Advanced maternal age (22.7%), abnormal ultrasound finding (38%), familial congenital disease (16%), abnormal fetal karyotype (5.6%), parental anxiety (17%), altered serum screening (0.7%) 269
Fiorentino et al (2011), prospective (cross-sectional) Italy Whole-genome BAC microarrays–CytoChip Focus Constitutional AF 938, CVS 99 Advanced maternal age 42.8%), altered serum screening (1.3%), abnormal ultrasound (4.6%), abnormal fetal karyotype (0.8%), family history of genetic disease (1.1%), parental anxiety (46.8%), others (2.4%) 1037
Wapner et al (2012), prospective (cross-sectional) United States Agilent 4-plex array and Affymetrix genomewide human SNP array 6.0 AF 1627, CVS 1910 Abnormal ultrasound finding (25.8%), advanced maternal age (47.9%), altered serum screening (19.3%), others (9.7%) 4282
Lee et al (2012), prospective (cross-sectional) Taiwan 1-Mb resolution BAC from 2010, until 60-K oligonucleotide AF 2926, CVS 82, fetal blood 93 Abnormal ultrasound findings (6.1%), altered serum screening (0.8%), advanced maternal age (60.2%), parental anxiety (31.1%) 3171
Armengol et al (2012), prospective Spain Not defined AF 728, CVS 164, fetal blood 14 Abnormal ultrasound finding (19%), altered serum screening (25.9%), history of congenital disease (16%), advanced maternal age (30.1%), parental anxiety (6.6%), other (2.2%) 906

AF , amniotic fluid; AMA , advanced maternal age (≥35 years old); BAC , bacterial artificial chromosome; CVS , chorionic villus sampling; SNP , single nucleotide polymorphism.

Saldarriaga. Karyotype vs genomic hybridization for prenatal diagnosis of chromosomal abnormalities. Am J Obstet Gynecol 2015 .


Characteristics of the excluded studies


The reasons for excluding these articles were as follows: postnatal diagnosis (13%), unrelated topic or outcome (60%), lack of comparison between the tests (6%), and lack of platform assessment (1.5%).


Risk of bias within the studies


For the studies of Armengol et al (2012), Lee et al (2012), Maya et al (2010), Van den Veyver et al (2009), and Wapner et al (2012), we observed an unclear risk of bias for the assessment of index and reference tests, mainly because the authors did not describe the blinding of the evaluation. The studies of Lee et al (2012) and Wapner et al (2012) exhibited an unclear risk of bias for the assessment of patient selection. There was only 1 study associated with a low risk of bias with respect to all items. ( Figure 2 ).




Figure 2


Risk of bias within studies

Saldarriaga. Karyotype vs genomic hybridization for prenatal diagnosis of chromosomal abnormalities. Am J Obstet Gynecol 2015 .


Risk of bias across the studies


Although we did not observe a high risk of bias, it is important to notice an unclear risk of bias for the index and reference test items. In addition, we observed a low risk of bias related to applicability concerns in all evaluated items ( Figure 3 ).




Figure 3


Risk of bias across studies

Saldarriaga. Karyotype vs genomic hybridization for prenatal diagnosis of chromosomal abnormalities. Am J Obstet Gynecol 2015 .


Results of individual studies


CGH vs gold standard (CGH + karyotype) including unknown significance variables


We observed a sensitivity of 94.5% (95% CI, 83.7–98.3%) and a specificity of 98.7% (95% CI, 97–99.4%) associated with high heterogeneity (I 2 = 84% and 81%, respectively).


The negative likelihood ratio was 0.032 (95% CI, 0.017–0.058) and the positive likelihood ratio was 71 (95% CI, 31–161) including a high heterogeneity (I 2 = 66-81%, respectively) ( Figures 4 and 5 , and Table 2 ).




Figure 4


CGH vs gold standard (CGH + karyotype): sensitivity and specificity forest plot

CGH , comparative genomic hybridization.

Saldarriaga. Karyotype vs genomic hybridization for prenatal diagnosis of chromosomal abnormalities. Am J Obstet Gynecol 2015 .



Figure 5


CGH vs gold standard (CGH + karyotype): NLR and PLR forest plot

CGH , comparative genomic hybridization; NLR , negative likelihood ratio; PLR , positive likelihood ratio.

Saldarriaga. Karyotype vs genomic hybridization for prenatal diagnosis of chromosomal abnormalities. Am J Obstet Gynecol 2015 .


Table 2

Summary of findings for comparative genomic hybridization and karyotype including unknown significance variants



























































































Item CGH (including USV) Karyotype (including USV)
Random effect analysis Random effect analysis
Negative likelihood ratio 0.032 0.291
95% CI 0.017–0.058 0.0841–1.011
P value < .001 .052
Heterogeneity ( P value) .02 .845
I 2 66% 0%
Positive likelihood ratio 71.898 866.365
95% CI 31.942–161.834 223.017–3365.650
P value < .001 < .001
Heterogeneity ( P value) < .001 .315
I 2 81% 16%
Sensitivity 0.945 0.673
95% CI 0.837–0.983 0351–0.886
P value < .001 .29
Heterogeneity ( P value) < .001 < .001
I 2 84% 96%
Specificity 0.987 0.99
95% CI 0.970–0.994 0.998–1
P value < .001 < .001
Heterogeneity ( P value) < .001 .637
I 2 81% 0%

CGH , comparative genomic hybridization; CI , confidence interval; USV , unknown significance variants.

Saldarriaga. Karyotype vs genomic hybridization for prenatal diagnosis of chromosomal abnormalities. Am J Obstet Gynecol 2015 .


Karyotype vs gold standard (CGH + karyotype) including unknown significance variables


We observed a sensitivity of 67.3% (95% CI, 35.1–88.6%) and a specificity of 99% (95% CI, 99.8–100%) associated with high (I 2 = 96%) and low (I 2 = 0%) heterogeneity, respectively.


The negative likelihood ratio was 0.29 (95% CI, 0.084–1.011) and the positive likelihood ratio was 866 (95% CI, 223–3365) associated with low heterogeneity (I 2 = 0-16%) ( Figures 6 and 7 , and Table 2 ).




Figure 6


Karyotype vs gold standard (CGH + karyotype): sensitivity and specificity forest plot

CGH , comparative genomic hybridization.

Saldarriaga. Karyotype vs genomic hybridization for prenatal diagnosis of chromosomal abnormalities. Am J Obstet Gynecol 2015 .



Figure 7


Karyotype vs gold standard (CGH + karyotype): NLR and PLR forest plot

CGH , comparative genomic hybridization; NLR , negative likelihood ratio; PLR , positive likelihood ratio.

Saldarriaga. Karyotype vs genomic hybridization for prenatal diagnosis of chromosomal abnormalities. Am J Obstet Gynecol 2015 .


Sensitivity analysis


We performed a sensitivity analysis for the inclusion or exclusion of unknown significance variables ( Table 3 ). With respect to CGH, we observed no differences in the sensitivity, specificity, or negative likelihood ratio. However, increases in the positive likelihood ratio and diagnostic odds ratios were associated with CGH. It is important to note that I 2 decreases for negative/positive likelihood ratios and specificity ( Table 3 ).


May 6, 2017 | Posted by in GYNECOLOGY | Comments Off on Karyotype versus genomic hybridization for the prenatal diagnosis of chromosomal abnormalities: a metaanalysis

Full access? Get Clinical Tree

Get Clinical Tree app for offline access