Patient choice and the randomized controlled trial
Article first published online: 23 MAY 2005
Copyright © 2005 ISUOG. Published by John Wiley & Sons, Ltd.
Ultrasound in Obstetrics & Gynecology
Volume 25, Issue 6, pages 535–536, June 2005
How to Cite
Hyett, J. (2005), Patient choice and the randomized controlled trial. Ultrasound Obstet Gynecol, 25: 535–536. doi: 10.1002/uog.1916
- Issue published online: 23 MAY 2005
- Article first published online: 23 MAY 2005
There are now many large series reporting the effectiveness of nuchal translucency (NT) screening for the detection of trisomy 21 in the literature, but to date there has been no randomized controlled trial comparing this screening option to more traditional approaches. The first large multicenter study reporting screening by a combination of maternal age (MA) and NT at 10–14 weeks' gestation demonstrated that 77% of trisomy 21 fetuses could be detected for a 5% false-positive rate1. This compared favorably with a program of screening by MA alone, which had previously been shown to have a 30% detection rate for the same 5% false-positive rate2. This, and subsequent interventional studies, have been criticized for inflating detection rates, as they include some trisomy 21 fetuses that would have been spontaneously lost during the course of pregnancy. In reality, the proportion of trisomic fetuses aborting between 12 and 16 weeks of pregnancy, when screening by MA or with biochemical markers occurs, is small, and these fetuses would not exclusively have had increased NT3, 4.
An alternative method of studying the effectiveness of NT screening is through an observational study where NT is assessed but not reported. These studies are, however, subject to their own bias, as there is frequently insufficient time and energy spent training staff to measure NT correctly and other clinical priorities may override the need to complete the scan. Consequently, observation trials frequently report higher failure rates of NT measurement and lower detection rates for NT screening. One observational study gave sonographers written instructions on how to measure NT but there was no practical training or attempt to standardize measurement. There were large variations between centers in measurement of NT and the sensitivity of screening; for a 5% false-positive rate, the detection rate was 31% (range 0–100%)5. In another study of 47 053 women, NT was not measured in 4228 patients as they did not attend at a suitable gestation, sonographers were unable to obtain a measurement in 3416 pregnancies and image review deemed a further 3881 scans to be of unacceptable quality6. Consequently, NT was only successfully measured in 77% of pregnancies and the sensitivity of the test was described as being poor (38% for 5% false-positive rate).
Randomized controlled trials have been shown to be the most rigorous way of comparing two treatment options and have had a significant impact on clinical practice through the development of evidence-based medicine7. This technique aims to remove systematic bias through randomization and, where feasible, blinding both clinicians and patients to the process. In this issue of the journal, Saltvedt et al. report the effect of screening 39 572 women randomized to trisomy 21 screening policies based either on NT assessment or on MA8. Women with a previous history of trisomy 21 or with other abnormalities noted at the time of the anomaly scan were also offered invasive testing. The trial aimed to determine whether NT screening led to a reduction in the number of liveborn infants with trisomy 21. A secondary outcome measure was the number of invasive tests performed to detect one trisomic fetus.
The trial design addressed the potential bias of ‘the learning curve’ by ensuring that all 72 midwives and doctors involved in the study were trained in the standardized assessment of NT prior to starting the study. As the women were recruited and randomized at the same gestation in both arms of the study, and the mean gestation of invasive testing was the same in the two groups, the bias of intrauterine lethality was minimized, although it was unfortunate that spontaneous fetal losses were not routinely karyotyped.
In the trial, NT detected 77% of trisomy 21 pregnancies for a false-positive rate of 5% and MA detected 58% of pregnancies for a false-positive rate of 18%. The same detection rates as described in previous interventional studies. There were 28% more trisomy 21 fetuses in the NT group, but this was not statistically significant. There was a significant difference in the proportion of liveborn trisomy 21 fetuses: 10 (18%) of 55 in the NT group and 16 (37%) of 43 in the MA group (χ2: P < 0.05), but no significant difference in the prevalence of trisomy 21 fetuses in all live births; 10 in 19 796 in the NT group vs. 16 in 19 776 in the MA group (χ2 test: P = 0.24). One trisomy 21 fetus was diagnosed for every 38 invasive tests performed in the NT group compared to 1 : 85 in the MA group.
On face value, the trial shows that although NT assessment improves the detection rate for trisomy 21 and has a better positive predictive value for screening, it does not reduce the live-birth prevalence of Down syndrome compared to more traditional screening techniques. The results are, however, compromised by the effect of bias introduced by patient choice. During recruitment, 20% of women considered to be eligible for the trial chose not to participate, despite the fact that they were not restricted to acting on the results of the screening they were randomized to, and they could also request and act on other screening tests (second-trimester serum biochemistry) if they wished. Only 32% of the invasive tests performed in the NT group were performed for an indication proposed in the study design; 47.4% were performed because of advanced MA. Whilst testing for a ‘true indication’ identified 39 of the trisomy 21 fetuses in this group, testing for MA identified none. If the groups had been karyotyped on the basis of the policies described, one trisomy 21 fetus would have been diagnosed for every 16 invasive tests performed in the NT group compared with 1 : 89 in the MA group. Similarly, although the majority of invasive tests performed in the MA group were for this primary indication (79.7%) and this led to the identification of 15 trisomy 21 fetuses, a further nine trisomic fetuses were detected on the basis of maternal serum screening or ultrasound.
Both physicians and the public express a wide range of opinions about Down syndrome screening. It is not surprising that many eligible patients did not want to participate in this study or that many women determined their own path for subsequent management rather than relying on the risk data generated by the trial protocol. It was not possible to conceal the screening allocation and the authors statement that the patients ‘did not trust’ NT in the Discussion shows that the patients, and possibly some clinicians, did not truly believe there was equipoise between the two screening techniques.
Defining the prevalence of liveborn trisomic infants as the primary outcome measure is attractive but is complicated by the fact that there are so many confounding variables between the time of screening and the time of birth. These factors make it difficult to assess attrition bias or to assess the data on an intention-to-treat basis and potentially leads to invalid conclusions about the efficacy of the two approaches to screening.
This well-designed randomized controlled trial once again demonstrates that NT assessment at 11–14 weeks' gestation is an effective tool in screening for trisomy 21. The reduction in the number of invasive tests performed in the process of prenatal diagnosis is important as this will lead to a significant reduction in miscarriages associated with these procedures. The results also show the difficulties of organizing such a complex study and we must bear these issues in mind when we consider their validity and relevance to our clinical practice.