BACKGROUND Common fetal aneuploidies include Down syndrome (trisomy 21 or T21), Edward syndrome (trisomy 18 or T18), Patau syndrome (trisomy 13 or T13), Turner syndrome (45,X), Klinefelter syndrome (47,XXY), Triple X syndrome (47,XXX) and 47,XYY syndrome (47,XYY). Prenatal screening for fetal aneuploidies is standard care in many countries, but current biochemical and ultrasound tests have high false negative and false positive rates. The discovery of fetal circulating cell-free DNA (ccfDNA) in maternal blood offers the potential for genomics-based non-invasive prenatal testing (gNIPT) as a more accurate screening method. Two approaches used for gNIPT are massively parallel shotgun sequencing (MPSS) and targeted massively parallel sequencing (TMPS). OBJECTIVES To evaluate and compare the diagnostic accuracy of MPSS and TMPS for gNIPT as a first-tier test in unselected populations of pregnant women undergoing aneuploidy screening or as a second-tier test in pregnant women considered to be high risk after first-tier screening for common fetal aneuploidies. The gNIPT results were confirmed by a reference standard such as fetal karyotype or neonatal clinical examination. SEARCH METHODS We searched 13 databases (including MEDLINE, Embase and Web of Science) from 1 January 2007 to 12 July 2016 without any language, search filter or publication type restrictions. We also screened reference lists of relevant full-text articles, websites of private prenatal diagnosis companies and conference abstracts. SELECTION CRITERIA Studies could include pregnant women of any age, ethnicity and gestational age with singleton or multifetal pregnancy. The women must have had a screening test for fetal aneuploidy by MPSS or TMPS and a reference standard such as fetal karyotype or medical records from birth. DATA COLLECTION AND ANALYSIS Two review authors independently carried out study selection, data extraction and quality assessment (using the QUADAS-2 tool). Where possible, hierarchical models or simpler alternatives were used for meta-analysis. MAIN RESULTS Sixty-five studies of 86,139 pregnant women (3141 aneuploids and 82,998 euploids) were included. No study was judged to be at low risk of bias across the four domains of the QUADAS-2 tool but applicability concerns were generally low. Of the 65 studies, 42 enrolled pregnant women at high risk, five recruited an unselected population and 18 recruited cohorts with a mix of prior risk of fetal aneuploidy. Among the 65 studies, 44 evaluated MPSS and 21 evaluated TMPS; of these, five studies also compared gNIPT with a traditional screening test (biochemical, ultrasound or both). Forty-six out of 65 studies (71%) reported gNIPT assay failure rate, which ranged between 0% and 25% for MPSS, and between 0.8% and 7.5% for TMPS.In the population of unselected pregnant women, MPSS was evaluated by only one study; the study assessed T21, T18 and T13. TMPS was assessed for T21 in four studies involving unselected cohorts; three of the studies also assessed T18 and 13. In pooled analyses (88 T21 cases, 22 T18 cases, eight T13 cases and 20,649 unaffected pregnancies (non T21, T18 and T13)), the clinical sensitivity (95% confidence interval (CI)) of TMPS was 99.2% (78.2% to 100%), 90.9% (70.0% to 97.7%) and 65.1% (9.16% to 97.2%) for T21, T18 and T13, respectively. The corresponding clinical specificity was above 99.9% for T21, T18 and T13.In high-risk populations, MPSS was assessed for T21, T18, T13 and 45,X in 30, 28, 20 and 12 studies, respectively. In pooled analyses (1048 T21 cases, 332 T18 cases, 128 T13 cases and 15,797 unaffected pregnancies), the clinical sensitivity (95% confidence interval (CI)) of MPSS was 99.7% (98.0% to 100%), 97.8% (92.5% to 99.4%), 95.8% (86.1% to 98.9%) and 91.7% (78.3% to 97.1%) for T21, T18, T13 and 45,X, respectively. The corresponding clinical specificities (95% CI) were 99.9% (99.8% to 100%), 99.9% (99.8% to 100%), 99.8% (99.8% to 99.9%) and 99.6% (98.9% to 99.8%). In this risk group, TMPS was assessed for T21, T18, T13 and 45,X in six, five, two and four studies. In pooled analyses (246 T21 cases, 112 T18 cases, 20 T13 cases and 4282 unaffected pregnancies), the clinical sensitivity (95% CI) of TMPS was 99.2% (96.8% to 99.8%), 98.2% (93.1% to 99.6%), 100% (83.9% to 100%) and 92.4% (84.1% to 96.5%) for T21, T18, T13 and 45,X respectively. The clinical specificities were above 100% for T21, T18 and T13 and 99.8% (98.3% to 100%) for 45,X. Indirect comparisons of MPSS and TMPS for T21, T18 and 45,X showed no statistical difference in clinical sensitivity, clinical specificity or both. Due to limited data, comparative meta-analysis of MPSS and TMPS was not possible for T13.We were unable to perform meta-analyses of gNIPT for 47,XXX, 47,XXY and 47,XYY because there were very few or no studies in one or more risk groups. AUTHORS' CONCLUSIONS These results show that MPSS and TMPS perform similarly in terms of clinical sensitivity and specificity for the detection of fetal T31, T18, T13 and sex chromosome aneuploidy (SCA). However, no study compared the two approaches head-to-head in the same cohort of patients. The accuracy of gNIPT as a prenatal screening test has been mainly evaluated as a second-tier screening test to identify pregnancies at very low risk of fetal aneuploidies (T21, T18 and T13), thus avoiding invasive procedures. Genomics-based non-invasive prenatal testing methods appear to be sensitive and highly specific for detection of fetal trisomies 21, 18 and 13 in high-risk populations. There is paucity of data on the accuracy of gNIPT as a first-tier aneuploidy screening test in a population of unselected pregnant women. With respect to the replacement of invasive tests, the performance of gNIPT observed in this review is not sufficient to replace current invasive diagnostic tests.We conclude that given the current data on the performance of gNIPT, invasive fetal karyotyping is still the required diagnostic approach to confirm the presence of a chromosomal abnormality prior to making irreversible decisions relative to the pregnancy outcome. However, most of the gNIPT studies were prone to bias, especially in terms of the selection of participants.