Second-Trimester Ultrasound to Detect Fetuses with Down Syndrome
Abstract & Commentary
In february 2001, an article by smith-bindman and colleagues appeared in the Journal of the American Medical Association that has generated substantial controversy. Smith-Bindman et al pooled data from 59 prospective, retrospective, nonrandomized studies in which ultrasound markers of Down syndrome (DS) were evaluated. These included choroid plexus cysts (CPC), nuchal skin-fold thickness (NSFT), echogenic intracardiac focus (EIF), echogenic bowel, and renal pyelectasis. In addition, both humeral and femur lengths were evaluated as predictors of DS. The meta-analysis involved 1930 pregnancies affected by DS and 130,363 unaffected pregnancies, chosen from studies that met Smith-Bindman et al’s seemingly stringent inclusion requirements.
After analyzing the jumble of data, Smith-Bindman et al con-cluded that isolated markers, with the possible exception of NSFT, had only a "marginal impact on the risk of DS." They even suggest-ed that clustered markers had little screening value since they increased the false-positive rate. Finally, they concluded that a "negative" sonogram did not substantially decrease the risk of DS.
In the discussion section of the paper, Smith-Bindman et al took their findings to an interesting conclusion—that if one were to act on isolated markers, an alarming number of normal fetuses would be lost through unnecessary amniocentesis.
The concept that noninvasive testing combinations that have less than 100% accuracy can be offered as an alternative to an invasive test has not been uniformly embraced by all obstetricians and geneticists. Smith-Bindman et al’s conclusions that the genetic sonogram is not only essentially useless, but may be dangerous, is music to the ears of the techniques’ naysayers. However, before obstetricians yield to the rallying cry of the "amniocentesis for all" faction, it is important to scrutinize the study data and Smith-Bindman et al’s conclusions. (Smith-Bindman R, et al. JAMA. 2001;285:1044-1055.)
Comment by John C. Hobbins, MD
First, a meta-analysis is designed to boost statistical power by combining data from many "similar" related studies, thus allowing conclusions to be reached that might not be supported by each of the individual studies. This technique has been especially effective in prospective, randomized, clinical trials (RCTs). However, this analysis has occasionally been used effectively in nonrandomized observational studies, as long as certain critical assumptions are verified and appropriate analytical safeguards are enforced.
The meta-analysis above contained many studies that were published before 1995, indicating that the data were collected in the late 1980s and early 1990s when investigators were first acquiring experience with the ultrasound techniques, often on vintage equipment. Including these early "learning stage" studies in the meta-analysis could skew the results. In fact, a table in the study demonstrates that the sensitivity is almost uniformly improved as the timing of the study approaches the late 1990s.
Also, rigid standardization of the ultrasound techniques (similar to that used in today’s first trimester nuchal translucency trials) is essential in the critical evaluation of the true value of any method. This was not the case in many studies included in the meta-analysis.
Smith-Bindman et al point out that a single ultrasound marker for DS has very little predictive value, but, hopefully, most clinicians are aware that 1 marker in isolation is not a reason to move to amniocentesis, especially in a patient at low risk for fetal DS. However, Smith-Bindman et al’s assertion that a "negative" sonogram has no merit is inconceivable since their own data indicate its’ substantial value. For example, their data summarized in one table shows that the sensitivity of multiple markers and/or the presence of structural abnormalities is 69% and the specificity is 92%. This results in a negative-likelihood ratio of 0.33 (100-69/92). Therefore, in patients with a negative sonogram, their risk could be adjusted downward from their age-adjusted risk by 66%. This assumption would take the DS risk of a 35-year-old to that of a 31-year-old, and certainly way below the risk of amniocentesis.
Vintzileos has pointed out that the ability to change a patient’s risk for DS (from her age-related risk) is based on the accuracy (sensitivity) of the center in which the ultrasound examination is performed. His center’s sensitivity is 82% with a specificity of 91%, which translates into a negative likelihood ratio of 0.20 (of a negative genetic sonogram). This, in turn, would drop a patient’s risk of fetal DS by 80%. Nyberg has published similar results, which are also similar to data from a large 9-center study soon to be submitted. Interestingly, the RR of a negative sonogram in this JAMA meta-analysis is extremely similar to this and other published data.
The thrust of the genetic sonogram has been to diminish the need for amniocentesis, especially in patients who would have been urged to have one a few years ago. What is the alternative? If we were to perform an amniocentesis on every patient who was 35 years old with a second trimester fetal DS risk of 1:300, we would be doing 299 unnecessary amniocentesis to identify 1 DS fetus, losing 2.4 fetuses along the way (using Smith-Bindman et al’s 0.8% amnio loss rate).
A genetic sonogram has a similar value to a "triple screen" which can also diminish the need for amniocentesis. As indicated in a previous clinical study, Egan et al calculated from 1997 national statistics that if triple screens were offered to all patients older than age 34, and amniocentesis was performed in only those with a risk of 1:250 or greater, 154,756 fewer amniocentesis would be performed and 733 procedure-related fetal losses would be averted.
It remains to be seen through expanded studies in progress whether the genetic sonogram and the triple screen will have greater adjunctive power than either alone, but initial findings suggest benefit when they are used together. A caveat must be added to any statement regarding the value of a genetic sonogram. Through bitter experience, it is clear that the benefit of this type of technique, in contrast to the mechanized triple screen, is dependent upon the skill and diligence of the operator and will suffer if rigid standardization is not pursued during the performance of the examination. In fact, many have suggested that this type of screening program would be best conducted in a high volume, high efficacy setting which would favor large-center screening programs.
In summary, a genetic sonogram should consist of a fetal anatomy survey (as described in the AIUM guidelines for a basic ultrasound examination), a femur and humerus length assessment, and an evaluation of DS markers. This should always include an assessment of the NSFT, which even in the meta-analysis increased the risk of DS by 17-fold, if positive. Every aspect of this exam must be compulsively standardized, and the exam should be performed by sonographers/sonologists with substantial experience. Based on data in the literature, if vigorous safeguards are used, one can diminish a patient’s risk of DS by at least 50% from her pre-exam risk.
Suggested Reading
1. Egan JF, et al. Obstet Gynecol. 2000;96:979-985.
2. Vintzileos AM, et al. Am J Obstet Gynecol. 1999; 181:1045-1048.
3. Nyberg DA, et al. J Ultrasound Med. 2000;19(4 suppl): S7-S8.
You have reached your article limit for the month. Subscribe now to access this article plus other member-only content.
- Award-winning Medical Content
- Latest Advances & Development in Medicine
- Unbiased Content