OBJECTIVE The aim of this study was to determine the diagnostic accuracy of comparative genomic hybridization (CGH) compared with karyotyping for the detection of numerical and structural chromosomal alterations in prenatal diagnosis. METHODS A metaanalysis was performed using searches of PubMed, EMBASE, CENTRAL, Cochrane Register of Diagnostic Test Accuracy Studies, Google Scholar, gray literature, and reference manuals. No language restriction was imposed. We included cross-sectional, cohort, and case-control studies published from January 1980 through March 2014 in the analysis. Studies of pregnant women who received chorionic villus biopsies, amniocentesis, or cordocentesis and then underwent CGH and karyotype analysis were included. Two independent reviewers assessed each study by title, abstract, and full text before its inclusion in the analysis. Methodological quality was assessed using QUADAS2, and a third reviewer resolved any disagreement. Conclusions were obtained through tests (sensitivity, specificity, and likelihood ratios) for the presence of numerical and structural chromosomal abnormalities. The reference used for these calculations was the presence of any abnormalities in either of the 2 tests (karyotype or CGH), although it should be noted that in most cases, the karyotyping test had a lower yield compared with CGH. Statistical analysis was performed in RevMan 5.2 and the OpenMeta[Analyst] program. RESULTS In all, 137 articles were found, and 6 were selected for inclusion in the systematic review. Five were included in the metaanalysis. According to the QUADAS2 analysis of methodology quality, there is an unclear risk for selection bias and reference and standard tests. In the other elements (flow, time, and applicability conditions), a low risk of bias was found. CGH findings were as follows: sensitivity 0.939 (95% confidence interval [CI], 0.838-0.979), I(2) = 82%; specificity 0.999 (95% CI, 0.998-1.000), I(2) = 0%; negative likelihood ratio 0.050 (95% CI, 0.015-0.173), I(2) = 0%; and positive likelihood ratio 1346.123 (95% CI, 389-4649), I(2) = 0%. Karyotype findings were as follows: sensitivity 0.626 (95% CI, 0.408-0.802), I(2) = 93%; specificity 0.999 (95% CI, 0.998-1.000), I(2) = 0%; negative likelihood ratio 0.351 (95% CI, 0.101-1.220), I(2) = 0%; and positive likelihood ratio 841 (95% CI, 226-3128), I(2) = 10%. CONCLUSIONS This systematic review provides evidence of the relative advantage of using CGH in the prenatal diagnosis of chromosomal and structural abnormalities over karyotyping, demonstrating significantly higher sensitivity with similar specificity.