Providing reliable prenatal screening performance estimates is critical for patient counseling and policy-making. Women who choose prenatal screening for aneuploidy are likely to be concerned not only with the common aneuploidies but with all causes of intellectual disability and serious birth defects. Sequential prenatal screening (combined serum and ultrasound testing) for aneuploidy detection commonly is offered as a primary screening test. Among women identified as screen positive, cell-free (cf)DNA has been added recently as a secondary, noninvasive screening option, before the consideration of invasive diagnostic testing (eg, amniocentesis and karyotype). With the anticipation of lower costs in the future, cfDNA might be an alternative to sequential screening in the general population. Sequential and cfDNA tests are both noninvasive, and both identify common aneuploidies. Screening via cfDNA detects more common chromosome abnormalities (eg, trisomy 21, sex trisomies). Sequential screening can identify other aneuploidies (eg, triploidy), as well as chromosome abnormalities associated with fetal structural abnormalities.
When the advantages and disadvantages of routine sequential screening with routine cfDNA screening are compared, one important measure is the proportion and severity of chromosome abnormalities identified. When reporting these detection rates, authors need to carefully consider the impact of multiple well-described biases. For women to make informed choices in situations of this type, determining reliable comparative performance estimates is crucial.
The Issue
Sequential prenatal screening (combined serum and ultrasound testing) for aneuploidy detection is now offered commonly. Women identified as screen-positive (high risk) are then offered invasive testing and definitive diagnosis. Since 2011, cell-free (cf)DNA sequencing has become available and is now recommended as an intermediate option for screen-positive women who decline invasive testing. With the anticipation of further cfDNA price decreases, it is timely to consider the benefits and harms of offering cfDNA for primary screening as an alternative to sequential screening. When faced with selecting a screening option, women will benefit by receiving reliable information for each test, including the detection and false-positive rates, availability, consequences of a positive or negative test, and associated harms. This debate focuses on the comparison of disorders identified and associated detection rates for sequential vs cfDNA screening in the general pregnancy population.
In a recent publication, Norton et al claim that prenatal screening based on a combination of serum and ultrasound measurements will identify more chromosome abnormalities than will next-generation sequencing of circulating cell-free (cf)DNA. This finding seems counterintuitive. On the basis of our analysis of this publication and previous relevant studies by this group, we conclude that the claim is incorrect and have identified methodologic and interpretive issues that merit consideration. This is not a new or isolated instance of a high-profile publication of a cohort study containing overestimated detection rates of Down syndrome. Determining an accurate detection rate based on observed cases in which karyotypes were not obtained on all patients requires paying strict attention to potential biases, especially relating to ascertainment. Screening recommendations from one professional society place emphasis on the need to account for ascertainment bias in reporting observed detection rates of Down syndrome. Detection rates for cfDNA testing generally are based on studies in which karyotypes were performed in all of the pregnancies.
- •
The 93% detection rate reported for Down syndrome serum/ultrasound screening is an overestimate. This rate was derived by the use of data from a cohort of 452,901 women enrolled in the California Screening Program. The 1184 screen-positive cases detected in the first or second trimesters were divided by the total number of Down syndrome cases identified (1184 screen positive + 91 screen negative). In that analysis, the authors correctly determined that 245 Down syndrome pregnancies could not be accounted for (using reliable projections of expected Down syndrome births based on the maternal age distribution) but then ignored this information. It is surprising that the authors did not take this finding into account in either the previous or present publication.
Determining a reliable detection rate from a cohort study requires addressing 2 well-described ascertainment biases: (1) the trimester of ascertainment bias is an important concern when screen-negative women with affected fetuses experience a high rate of undetected spontaneous loss (45% for Down syndrome between the late first trimester and term). This source of bias is less serious with screen-positive cases, because affected pregnancies usually are identified early, via diagnostic testing. (2) Ascertainment of live born Down syndrome cases is also an important concern, especially because identifying all cases among women screened in a large prospective cohort study is difficult. The Figure shows how not accounting for these 2 biases will cause overestimation of the detection rate. On the basis of the 21% underascertainment of cases reported in California, the true detection rate of Down syndrome is more likely to be 80−85%. Not accounting for these 2 biases also will result in the authors overestimating the detection rates for trisomies 18 and 13.
- •
The 63% detection rate for sex chromosome trisomies by serum/ultrasound screening is a gross overestimate. Only recently has the detection rate for sex trisomies via the use of serum/ultrasound markers been explored. As with Down syndrome, the detection rate for sex chromosome trisomies in the California study was computed with only the screen-positive and screen-negative cases that had been identified during pregnancy or in the first year of life. Unlike the high fetal loss rates documented for autosomal trisomies, spontaneous fetal loss rates for sex chromosome trisomies are low, so the trimester of ascertainment bias is not an important issue. Instead, incomplete ascertainment among live births is a far more important potential source of bias. The authors state in the methods that “infants with no karyotyping performed in their first year were presumed to be euploid.”
