First-trimester ductus venosus screening for cardiac defects: a meta-analysis

Authors


JPA Ioannidis, Stanford Prevention Research Center, Stanford University School of Medicine, 251 Campus Drive, MSOB X306, Stanford, CA 94305, USA. Email jioannid@stanford.edu

Abstract

Please cite this paper as: Papatheodorou S, Evangelou E, Makrydimas G, Ioannidis J. First-trimester ductus venosus screening for cardiac defects: a meta-analysis. BJOG 2011;118:1438–1445.

Background  Heart defects are the most common congenital abnormalities.

Objective  We aimed to evaluate in a meta-analysis the screening performance of abnormal ductus venosus (DV) Doppler waveform for detection of congenital heart disease (CHD) in chromosomally normal fetuses.

Search strategy  Studies were retrieved from a search of MEDLINE, ISI, SCOPUS and EMBASE (from 1999 to March 2011) using the keywords ‘ductus venosus’, ‘DV’, ‘chromosomal abnormalities’, ‘congenital heart disease’ and ‘nuchal translucency’.

Selection criteria  We considered all studies that examined the diagnostic performance of DV in the first trimester for CHD in chromosomally normal fetuses. We included studies that were limited to fetuses with increased nuchal translucency (NT), normal NT, and studies that examined fetuses regardless of NT status.

Data collection and analysis  Seven studies (n = 50 354) regardless of the NT status, nine studies (n = 2908) with increased NT and seven studies (n = 47 610) with normal NT were included in the meta-analysis. We drew hierarchical summary receiver operating characteristic (HSROC) curves using the parameters of the fitted models.

Main results  In populations including participants regardless of NT status, the summary sensitivity and specificity of DV for detecting CHD were 50 and 93%, respectively. In participants with increased NT, the summary sensitivity and specificity were 83 and 80%, and in those with normal NT, they were 19 and 96%, respectively.

Authors’ conclusions  The estimated performance of DV assessment for detection of CHD in chromosomally normal fetuses can be considered in evaluating the potential use and limitations of this screening test.

Introduction

Heart defects and abnormalities of the great vessels are the most common congenital malformations and they are found in between two and eight of every 1000 pregnancies.1 Ultrasound markers have been increasingly used to detect congenital heart disease (CHD) as early as the first trimester. The best established marker for first-trimester screening, increased nuchal translucency (NT), has been robustly associated with cardiac dysfunction, with or without structural anomalies, in chromosomally normal and abnormal fetuses.2–7 However, increased NT has only modest sensitivity and specificity. Many fetuses with increased NT may have no CHD and may undergo unnecessary detailed cardiac scans. Moreover, some fetuses with CHD and normal karyotype are missed based on NT evaluation alone. Therefore there is interest in assessing whether other ultrasound markers may improve diagnostic discrimination. Abnormal flow in the ductus venosus (DV) has been associated with adverse perinatal outcome,8,9 chromosomal abnormalities10,11 and CHD.12,13 Fetuses with abnormal karyotype have a high incidence of cardiac defects. Conversely, for fetuses with normal karyotype, there are several studies that examine the role of DV waveform abnormalities in the detection of CHD, but the results are variable concerning the ability of this marker to offer discriminating information in fetuses with increased NT and in those with normal NT.

The aim of this study is to evaluate systematically the diagnostic performance of DV waveform in screening for CHD in fetuses with normal karyotype. We analysed the available evidence in an inclusive meta-analysis taking NT measurements into account as well. This allowed us to synthesise information from a substantial number of studies and arrive at more robust and precise estimates of diagnostic performance of the DV waveform in the general population, in fetuses with high NT, and in those with normal NT.

Methods

Search strategy and eligibility criteria

Studies were retrieved from a search of MEDLINE, ISI, SCOPUS and EMBASE (from 1999 to March 2011) using the keywords ‘ductus venosus’, ‘DV’, ‘chromosomal abnormalities’, ‘congenital heart disease’ and ‘nuchal translucency’. The references of each eligible study were screened for possible missing articles. To avoid overlapping populations, in cases where participants were included in more than one publication, we include the one with the largest sample size.

Studies were eligible if they provided data on the presence or not of CHD according to the DV waveform during the first trimester in chromosomally normal fetuses. We considered studies that were limited to fetuses with increased NT, studies that addressed fetuses with normal NT, and studies that addressed fetuses regardless of NT status. When data were not provided separately for fetuses with increased NT and with normal NT, we tried to contact the investigators for additional information. Moreover, for a study to be eligible, DV and NT measurements had to be performed in the first trimester by experienced ultrasonographers according to the Fetal Medicine Foundation guidelines.14,15 We included studies where the fetal karyotype was determined either by chorionic villus sampling or amniocentesis and CHD was diagnosed by fetal echocardiography performed by specialists, post-mortem examination (in cases of fetal death) or postnatal physical examination.

Data extraction

For each study we recorded the name of the author, country of origin, sample size, high risk or unselected population, cut-off for the measurement of NT, study design, maternal age, gestational age at which the measurement was performed, mean crown–rump length (CRL) and outcomes. The prevalence of CHD in fetuses with normal karyotype was also calculated for each study. We also recorded the number of true positives, true negatives, false positives and false negatives for the presence or not of a cardiac defect according to normal or abnormal DV waveform pattern. Data were extracted by two independent investigators (SP, EE). A third investigator (JPAI) settled any remaining discrepancies and consensus was reached for all data.

Analysis

We used a bivariate model to obtain an overall sensitivity and overall specificity. We fitted a two-level mixed logistic regression model conditional on the sensitivity and the specificity of each study and a bivariate normal model for the sensitivity and specificity between studies.16 We also drew the hierarchical summary receiver operating characteristic (HSROC) curve based on the parameters of the fitted model. The HSROC curve shows the summary trade-off between sensitivity and specificity across the included studies.

We also calculated likelihood ratios for each study and summaries thereof across all studies. These metrics also combine both sensitivity and specificity in their calculation. Positive likelihood ratio (LR+) is the ratio of sensitivity over (1−specificity), whereas negative likelihood ratio (LR−) is defined as the ratio of (1−sensitivity) over specificity. When there is absolutely no discriminating ability for a diagnostic test, both likelihood ratios equal 1. The discriminating ability is better with higher LR+ and lower LR−. A good diagnostic test typically has LR+ greater than 5.0 and LR− less than 0.2.

We examined the overall diagnostic accuracy of DV in the whole population with normal karyotype (regardless of NT status) and separately in the population of fetuses with increased NT and normal karyotype and in the population of normal karyotype and normal NT.

Analyses were conducted in Intercooled Stata version 10 (Stata Corp, College Station, TX, USA).

Results

Eligible studies

Of the 967 items retrieved with the electronic search, 800 were excluded based on the title and the abstract. The remaining 167 articles were retrieved for screening in full text. We further excluded studies that examined DV Doppler during the second trimester, studies that did not include fetuses with CHD, case reports, meeting abstracts, editorials and studies that addressed specific technical issues about Doppler examination. Hence, 17 studies were eligible for analysis.12,13,17–31 Three studies were excluded because of overlapping populations.23,24,31 One article26 among the 14 that were potentially eligible had a case–control design and did not provide sufficient data on the chromosomal status of the fetuses and NT status; we contacted the investigators, but could not retrieve this information, therefore that study could not be included in the meta-analysis. In two studies, there were missing data that could not be retrieved despite conducting the authors.19,21 In another study, not all fetuses at high risk for chromosomal abnormalities underwent karyotype evaluation and we could not be certain about the chromosomal status of fetuses with CHD.25 Finally, one publication evaluated whether DV flow alterations in fetuses with increased NT are correlated with only a certain type of cardiac defect so it was excluded.20 Hence the final data included information from nine populations with increased NT, seven populations with normal NT, and seven populations including both fetuses with normal and increased NT (Figure 1).

Figure 1.

 Flow chart of the search strategy and selected studies.

Study characteristics

Descriptive characteristics of the eligible studies are presented in Table 1. For analysis of the diagnostic performance of DV regardless of the NT status seven studies contributed 50 354 fetuses. In total, nine studies with 2908 fetuses were included in the meta-analysis of populations with increased NT. Seven studies with 47 610 fetuses were available in populations with normal NT.

Table 1.   Study characteristics
AuthorCountryDesignNPrevalence of CHDPopulationMaternal ageGestational ageCRLMeasureCutoff
  1. DV, ductus venosus; NP, not provided; NT, nuchal translucency.

Favre et al.13FranceProspective cohort99810/998 (10‰)High risk (advanced maternal age, previous fetus with aneuploidy, increased NT)31.9 (16–46)12.7 (11–14)69 (40–91)DVAbsent or reversed flow during a-wave
NT>95th percentile
Haak et al.28The NetherlandsProspective cohort222/22 (9%)High risk (increased NT)32.7 (24–42)12.5 (11–14)NPDVIncreased PI and absent or reversed a-wave
NT>95th centile
Chelemen et al.17UKProspective cohort40 99085/40 990 (2‰)General population31 (14–51)12 (11–13)NP (45–84)DVAbsent or reversed flow during a-wave
NT>95th centile
Timmerman et al.18The NetherlandsProspective cohort79235/792 (4%)General population35 (19–46)13 (11–14)61 (40–86)DVIncreased PI and absent or reversed flow during a-wave
NT>95th centile
Toyama et al.27BrazilProspective cohort10757/1075 (6‰)General population32.1 (14–47)12 (11–14)65.3 (45–84)DVAbsent or reversed flow during a-wave
NT>95th centile
Zoppi et al.30ItalyProspective cohort2962/296 (7‰)High risk (exposure to teraatogenic agents, maternal anxiety, general fetal risks)35 (17–46)11.5 (10.3–13.6)50.5 (38–82)DVAbsent or reversed flow during a-wave
NT>95th centile
Matias et al.12UK, PortugalProspective cohort1427/142 (5%)High risk (increased NT)NP12 (10–14)NPDVAbsent or reversed flow during a-wave
NT>95th centile
Martinez et al.22SpainProspective cohort590645/5906 (7‰)General population33 (17–45)12 (11–14)59 (45–85)DVAbsent or reversed flow during a-wave
NT>99th centile
Murta et al29USAProspective cohort3431/343 (3‰)General population32 (17–47)12 (11–14)59 (38–84)DVAbsent or reversed flow during a-wave
NT>95th centile

The studies had been performed in Europe, USA and Brazil. All of them were prospective cohorts and four of them addressed high-risk populations, while the other five evaluated largely general population samples (Martinez et al.,22 Toyama et al.,27 Timmerman et al.18 and Chelemen et al.17 evaluated general populations, whereas Murta et al.22 examined 372 fetuses without increased risk for chromosomal abnormalities and included also 39 with increased NT). The sample size varied widely between the studies (22–40 990). The mean maternal age ranged from 31.9 to 35 years. Mean gestational age at the time of examination ranged from 12 to 12.7 weeks. Mean CRL ranged from 50.5 to 69 mm. The Doppler examination of the DV considered to be abnormal when there was increased Pulse Index or absent/reversed flow during the a-wave. Increased NT measurements were those above the 95th centile except for one study that described increased NT in those above the 99th centile.22 The sensitivity and specificity of the included studies for each population are shown in Figure 2. In our total meta-analysis database, about half of the examples of CHD were in the increased NT population (n = 101) whereas the other half were in fetuses with normal NT (n = 93).

Figure 2.

 Forest plot of sensitivity and specificity of DV for detecting CHD with the 95% CI for each population of the included studies.

Quality assessment

We used the QUADAS tool for quality assessment of the included studies32 (Table 2). Generally, the included studies meet most of the quality criteria of this checklist. However, in most of the studies it is not clear whether the results of the DV and NT measurements were known at the time of the cardiac scan. Also, the detailed description or a reference on how the specialised scan is performed is missing from most of the studies.

Table 2.   The QUADAS tool for quality assessment of the included studies
AuthorItem 1Item 2Item 3Item 4Item 5Item 6Item 7Item 8Item 9Item 10Item 11Item 12Item 13Item 14
  1. U, unclear; Y, yes.

  2. Item 1: Representative spectrum of patients; Item 2: Selection criteria clearly described; Item 3: Reference standard correctly classifies the target condition; Item 4: Time period short enough between reference standard and index test; Item 5: All or a random sample received verification; Item 6: Patients receive the same reference standard; Item 7: Reference standard independent of index test; Item 8: Index test described in sufficient detail; Item 9: Reference standard described in sufficient detail; Item 10: Index test interpreted independently; Item 11: Reference standard interpreted independently; Item 12: Same clinical data available when test results were interpreted; Item 13: Intermediate test reported; Item 14: Withdrawals of the study explained.

Favre et al.13YYYYYYYYYYUYYY
Haak et al.28YYYYYYYYYYUYYY
Toyama et al.27YYYYYYYYNYUYYY
Zoppi et al.30YYYYUYYYYYUYYY
Matias et al.12YYYYYYYYNYUYYY
Murta et al.29YYYYUUYYNYUYYY
Timmerman et al.18YYYYYYYYNYUYYY
Martinez et al.22YYYYYYYYYYYYYY
Chelemen et al.17YYYYYYYYYYYYYY

Data synthesis

For the normal karyotype population regardless of the NT status, the overall sensitivity of abnormal flow in DV to detect CHD was 0.50 (95% CI 0.27–0.73) whereas the specificity was 0.93 (95% CI 0.88–0.96). The LR+ of the test was 8.1 (95% CI 4.5–14.7) whereas the LR− was 0.52 (95% CI 0.32–0.85). The HSROC curve is shown in Figure 3.

Figure 3.

 HSROC curve for the diagnostic performance of DV for CHD in population including karyotype-normal fetuses regardless of nuchal translucency status. Each study is shown by a circle with size proportional to the weight of the evidence. The summary sensitivity and specificity are shown with bold squares and the 95% confidence region is also plotted.

The overall sensitivity of abnormal flow in DV in populations with increased NT was 0.83 (95% CI 0.51–0.95), whereas the overall specificity was 0.80 (95% CI 0.56–0.93) The LR+ of the test was 4.35 (95% CI 1.5–12.1), whereas the LR− was 0.20 (95%: 0.05–0.79). These results are shown in the HSROC curve in Figure 4.

Figure 4.

 HSROC curve for the diagnostic performance of DV for CHD in populations with increased nuchal translucency. Each study is shown by a circle with size proportional to the weight of the evidence. The summary sensitivity and specificity are shown with bold squares and the 95% confidence region is also plotted.

Of the seven studies that provided data on populations with normal NT, two had no fetuses with CHD and the other five had 1–55 fetuses with CHD. The overall sensitivity for DV in populations with increased NT was 0.19 (95% CI 0.12–0.29) whereas the specificity was 0.96 (95% CI 0.92–0.98). The LR+ was 4.97 (95% CI 2.3–10.8) and the LR− was 0.8 (95% CI 0.75–0.93). These results are shown in an HSROC curve in Figure 5.

Figure 5.

 HSROC curve for the diagnostic performance of DV for CHD in populations with normal nuchal translucency. Each study is shown by a circle with size proportional to the weight of the evidence. The summary sensitivity and specificity are shown with bold squares and the 95% confidence region is also plotted.

We were not able to conduct an analysis using different NT cut-offs because most of the studies did not give separate data for NT >95th and NT >99th centiles.

Discussion

The findings of this meta-analysis on chromosomally normal fetuses demonstrate that the DV waveform examination has moderate sensitivity for detecting CHD. In fetuses with normal karyotype DV has a sensitivity of 50% and a specificity of 93%. In fetuses with increased NT, sensitivity is much higher (83%) and still only about one in five of fetuses that have no CHD but that have increased NT will also show an abnormal DV waveform. In fetuses with normal NT, sensitivity is low (19%), while approximately 96% of fetuses with normal NT and no CHD will also have a normal DV waveform.

The International Society of Ultrasound in Obstetrics and Gynecology proposes fetal echocardiography at the time of detection of increased NT (11–14 weeks) if the NT is >3.5 mm with a second scan at 20–22 weeks of gestation.33 This recommendation is a realistic policy because generally there are not sufficient resources to perform detailed cardiac scans in all fetuses with NT >95th percentile. The exact use of DV for detecting CHD has to be considered on a case-by-case basis, taking into account whether the mother wishes to decrease the chances that she would need to perform additional diagnostic procedures or decrease the chances of missing a diagnosis of CHD. Screening by increased NT plus DV (using DV information in fetuses with already documented increased NT) rather than increased NT alone would miss about a sixth of the CHD diagnoses identified by increased NT, but would also decrease to one-fifth the proportion of fetuses that would need to undergo further screening for CHD. When NT is normal, then DV would pick about 20% of the instances of CHD missed by NT and would lead to further screening for CHD about 4% of the screened population. Screening by DV without using NT in the general population would only find about half of the CHD, therefore it cannot be recommended as an efficient strategy, even more so because NT has already an established, well-supported role in CHD screening.

In specialised centres, DV is used more as an additional examination in the screening for Down syndrome, specifically in those cases where initial screening with NT and first-trimester biochemical markers show a high risk for Down syndrome. The Fetal Medicine Foundation recommends examining fetuses with DV examination having an intermediate estimated risk for trisomy 21 (1:100 to 1:1000), including approximately 16% of the population.

Screening fetuses with NT and DV waveform has the convenience that it can be performed at the same time during the first trimester on a routine setting provided the sonographers are sufficiently experienced to recognise DV abnormalities. It is important to take into consideration that competence in DV Doppler velocimetry requires extensive supervised training. A specialist in fetal sonography needs to perform at least 80 sonograms until they reach appropriate level of training.15

The quality assessment showed that the included studies had overall reasonably good methodology. This offers some relative reassurance that bias may not have influenced the results substantially.

Our study has some limitations. There is considerable variation between the individual studies in the number of cases, total sample size and proportion of CHD cases. However, the bivariate meta-analysis and HSROC analyses take explicitly this diversity into account and can accommodate studies with populations of different risks, and different definition thresholds. We were not able to obtain missing data from some studies because the authors did not respond to us or because they did not get approval from the hospital ethics committee to provide such data. However, the collected databases are quite complete, especially for fetuses with increased NT. For populations of fetuses with normal NT, it would be useful to assemble additional information, so as to obtain a less uncertain estimate about the sensitivity of NT in this setting. The majority of the studies defined abnormal DV as either increased Pulse Index or absent/reversed a-wave so it is not possible to juxtapose the diagnostic performance of DV based on these two definitions in the same studies and across a larger meta-analysis. Finally, all analyses were limited explicitly to fetuses with normal karyotype, which is not known at the time of ultrasound evaluation in the first trimester. However, previous studies have traditionally preferred to target normal karyotype populations. The reason is that up to 75% of trisomic fetuses have increased NT and abnormal DV,29 and such fetuses would be recognisable also from other aspects including biochemical and other sonographic markers, while most of them succumb to spontaneous or—more often—iatrogenically induced miscarriage during the first or second trimester.

Conclusion

The estimated performance of DV assessment for the detection of CHD in chromosomally normal fetuses can be considered in evaluating the potential use and limitations of this screening test. Taking into account the fact that very few referral centres have the facilities to provide detailed cardiac scans, we should be careful about the interpretation of this screening test’s results and the information we give to the parents. Most of the instances of CHD are currently still diagnosed in the second or even the beginning of the third trimester34 and decisions for management are more difficult at this point. It is very important to find noninvasive diagnostic tests that will place safely a fetus from the low-risk group for CHD into a high-risk group or the opposite during the first trimester so that the therapeutic approach and the decision-making will occur as soon as possible, minimising maternal and fetal morbidity.

Disclosure of interests

There was no conflict of interest for any author.

Contribution to authorship

The original idea was proposed by GM. All four authors worked on developing the protocol. Data were collected by SP and EE with help from JPAI and GM. SP and EE conducted the analyses. All authors interpreted the results; SP and JPAI wrote the paper, GM and EE further edited it. All authors approved the final version.

Details of ethics approval

Not required as this is a meta-analysis of published manuscripts.

Funding

None.

Acknowledgement

None.

Ancillary