How accurate is a second-trimester prenatal ultrasound in the diagnosis of Down syndrome?

BACKGROUND: Several different ultrasonographic markers have been associated with Down syndrome. Although screening for fetal abnormalities associated with Down syndrome in the second trimester has become common, the accuracy of these markers is unknown.

POPULATION STUDIED: The meta-analysis included data on 130,365 unaffected fetuses and 1930 fetuses documented to have Down syndrome from 56 studies evaluating ultrasonographic markers for Down syndrome. These studies were published between January 1980 and February 1999. The mean maternal age was 34 years, and 88% of the studies looked at women with an increased risk of chromosomal abnormality.

STUDY DESIGN AND VALIDITY: The authors included English-language studies evaluating second-trimester ultrasound to detect Down syndrome (found by a MEDLINE search) and a review of bibliographies. Ninety-five percent of the studies used chromosomal analysis as the gold standard to assess for Down syndrome. Retrospective trials were included if the ultrasound interpretation and the diagnosis of Down syndrome were independently determined. Of 220 studies initially identified, 56 remained after poor-quality studies and studies without extractable data were excluded. The validity of the remaining studies was not explicitly described, although 2 reviewers independently evaluated them. The sensitivity and specificity for each marker were extracted and evaluated. Markers included choroid plexus cysts, nuchal fold thickening, echogenic intracardiac focus, echogenic bowel, renal pyelectasis, femur shortening, and humerus shortening. Although the individual studies used different definitions of these markers, more than 90% of the data could be combined by using standard definitions in the meta-analysis. Publication bias was not addressed, but given the large numbers involved in the study, there would need to be a large number of unpublished studies to affect the results. Pooled sensitivity, specificity, and positive and negative predictive values and likelihood ratios were calculated using averages weighted for sample size. The authors evaluated positive and negative predictive values using 2 prevalences, first at 1:700 (the overall prevalence of Down syndrome) and again at 1:300 (the prevalence of Down syndrome in babies born to women older than 35 years).

OUTCOMES MEASURED: The authors classified each pregnancy into 1 of 3 outcomes: unaffected, Down syndrome, or any chromosomal abnormality. The outcome of all chromosomal abnormalities was used primarily for studies evaluating choroid plexus cysts, where 50% of the studies did not include data specifically for Down syndrome.

RESULTS: The overall prevalence of Down syndrome in the analysis was 1.5%, much higher than the 0.1% overall population risk. The sensitivity of each marker as an isolated abnormality was low (1%-16%), while the specificities were all greater than 95%. The most accurate test was a thickened nuchal fold, for which 2% of fetuses at average risk with this finding would have Down syndrome (the positive predictive value). This was the only test in patients at average risk for Down syndrome for which more cases would be identified than fetuses lost because of complications of the amniocentesis. Depending on the marker used, between 4454 and 87,413 average-risk women would have to be screened to detect 1 case of Down syndrome. For high-risk patients, between 1911 and 37,500 would be need to be screened. False-positives would range from 79 to 611, which means following most markers with an amniocentesis would result in more fetal loss than diagnoses of the disease. Because of the rarity of the disease, the absence of markers provides no significant reduction in the risk of having an infant with Down syndrome.