Should we implement universal screening with late pregnancy ultrasound to prevent stillbirth?

The first description of ultrasound to estimate fetal weight was published in 1975.1 More than 40 years later, we still do not know whether the universal use of this method prevents perinatal deaths. Multiple lines of evidence suggest that it should work: (1) stillbirth is strongly associated with fetal growth restriction (FGR);2 (2) current methods using targeted ultrasound have low detection rates for FGR;3 (3) there is strong evidence that the induction of labour at term reduces perinatal mortality.4 Despite this, the National Institute for Health and Care Excellence (NICE) Guideline on Antenatal Care specifically recommends against universal ultrasound.5 The basis for this view is a Cochrane review of randomised controlled trials (RCTs), which fails to show any benefit to the mother or the baby with universal scanning.6 An argument can be made that we should just go ahead and implement late pregnancy ultrasound anyway. First, the Cochrane review is imperfect. The definition of screen positive differed among the 13 trials in the review. The one thing that they all had in common was that none coupled the screening test with a proven, disease-modifying intervention. Furthermore, even if all the trials had been perfect, the total number of women in the meta-analysis, 35 000, is well below that required based on sample size calculations, making optimistic assumptions of both the diagnostic accuracy of ultrasound and the clinical effectiveness of the intervention.7 Stillbirth is a devastating complication in pregnancy, and rates of stillbirth are falling far more slowly than for neonatal death. Can't we do anything? Surely we must do something? But the question is whether universal late pregnancy ultrasound is the right thing to do. Just about every issue that concerns the outcome for the infant in obstetrics has to be regarded in the light of gestational age. Preterm birth is one of the strongest predictors for adverse outcome for the baby in both the short and the long term: the more preterm the birth, the stronger the associations with adverse outcome. The primary method that we have for preventing stillbirth is to deliver the baby. Much of maternal–fetal medicine depends on balancing the risk of neonatal morbidity and mortality if we deliver the baby versus the risk of intrauterine fetal death if we do not. Given that one-third of all stillbirths occur at term, and given that preterm birth is the major concern about intervention, it follows that screening to prevent stillbirth at term would be a logical first place to start. Although there is a view that FGR is primarily a problem of preterm births, small for gestational age (SGA) infants have higher risks of stillbirth at term: babies weighing below the third percentile were eight times more likely to be stillbirths at term.2 If we could detect these fetuses at around 36 weeks of gestation, could we not then follow them up closely with antepartum fetal surveillance coupled with the relatively benign intervention of early delivery in the event of any concerns? One obstacle to considering this question identified in the 2008 NICE Guideline on Antenatal Care was the lack of high-quality studies of the diagnostic effectiveness of ultrasound as a screening test for SGA, and the guideline recommended that such studies should be performed.5 Level-1 studies of diagnostic effectiveness (the equivalent of the double-blind, placebo-controlled randomised trial of interventional studies) involve blinding to the result of the new test. In 2015, we reported such a study in The Lancet and demonstrated that universal ultrasound is a good screening test for SGA in unselected nulliparous women with a singleton pregnancy.3 Taking the example of detecting severe SGA, i.e. birthweights below the third percentile, the area under the receiver operating characteristic (ROC) curve for an ultrasonic estimated fetal weight (EFW) at 36 weeks of gestation was 0.91. Defining screen positive as an EFW below the tenth percentile, selective ultrasound only detected 32% of these babies, whereas universal ultrasound increased detection to 77%, with a positive likelihood ratio of 6.1. There were similarly encouraging results when the outcome was all SGA, i.e. birthweights below the tenth percentile. Given these findings, given that we know that small babies are at an increased risk of stillbirth, and given that induction of labour at term reduces the risk of perinatal death by 70%,4 should we now proceed to implement screening? There are two main issues that should restrain our enthusiasm. The first is that, in the context of screening for stillbirth at term, a good screening test for SGA may not be not enough. The risk of antepartum stillbirth not caused by congenital anomaly in about a million term births in Scotland was 1.5 per 1000 births.2 Only about 30% of these were associated with birthweights outside the range of the 20–97th percentiles. Hence, the risk of term stillbirth associated with abnormal birthweight (using a broad definition) is about 1 in 2000. The risk would be even lower if confined to losses with birthweights below the third percentile. Hence, even if a screening test had a positive likelihood ratio of 100 for this outcome, the absolute risk would still be <5%. But with ultrasound we have a test with a positive likelihood ratio of ~6.0 for the outcome of a birthweight below the third percentile, where the vast majority of babies are liveborn. It would be impossible to conduct a study that was sufficiently powered to assess the diagnostic effectiveness of universal ultrasound as a predictor of term stillbirth associated with abnormal growth. On the basis of the available evidence, however, it is likely that only a minority of term stillbirths would have an EFW below the tenth percentile at 36 weeks of gestation, and hence universal ultrasound in late pregnancy coupled with the early delivery of SGA fetuses would have a relatively modest effect on the overall stillbirth rates. It would, however, have the potential for widespread negative effects. The Pregnancy Outcome Prediction (POP) study demonstrated that universal ultrasound had a specificity of 87%, i.e. 13% of normally grown fetuses would have been classified by ultrasound as SGA at 36 weeks of gestation.3 Making this diagnosis, or rather misdiagnosis, would have multiple adverse impacts. First, 13% of women with a non-SGA infant would be incorrectly told that there is a problem with fetal growth, creating anxiety and fear. Second, resources would then be used to manage their newly (and wrongly) defined high-risk pregnancies. Third, being told of the risks, they may opt for early term delivery with the aim of preventing stillbirth. Multiple studies show increased rates of short-term morbidity with early term delivery (at 37–38 weeks of gestation), and long-term follow up shows higher risks of special educational needs at school.8, 9 Hence, the consequence of false-positive results could be a substantial increase in short- and long-term complications for 10–15% of the population of healthy pregnant women. The POP study indicated that combining ultrasonic SGA with the best marker of FGR (reduced growth velocity) reduced the screen-positive rate to 4.3%. Only 12% (4/33) of cases of severe adverse perinatal outcome exhibited this combination in the last screening scan prior to delivery, however, reflecting two facts: (1) ultrasound has limited capacity to screen for FGR; and (2) many cases of severe adverse outcome are unrelated to FGR. Is there any evidence to support this gloomy assessment? A careful analysis of data from the 2010 French National Perinatal Survey supports the capacity for harm with universal ultrasound. A late pregnancy scan was implemented at 30–34 weeks of gestation in France. Monier and colleagues found that, even among infants for which screening correctly identified as SGA, there was no clear benefit with prenatal detection, indeed some adverse outcomes were more common in the true positives than in the false negatives.10 But what was particularly concerning in the context of ad hoc population screening was the much higher rates of adverse outcomes among the false positives than among the true negatives: i.e. among otherwise low-risk women who were incorrectly identified as having an SGA fetus, the risk of preterm birth was three-fold, the need for neonatal resuscitation was six-fold and the risk of admission to neonatal intensive care was doubled. Although the French approach involved screening in the preterm period rather than immediately prior to term, there is still the same potential for harm with a scan at 36 weeks of gestation, given the higher rates of complications with early term birth.8 Screening in the UK is strictly regulated. The National Screening Committee assesses the evidence and makes recommendations. The assessment involves far more than just the availability of a possible screening test and a possible intervention. It requires us to know the natural history of a positive test result in an unselected population, and the result of any associated intervention in the women who screen positive. Moreover, even if screening was proven to be safe and effective, some women may prefer not to be screened. The ability of women to make an informed choice depends on the availability of high-quality information on the risks and benefits of screening, as has been developed for breast cancer screening (https://www.gov.uk/government/publications/breast-screening-helping-women-decide). So how do we move forwards? In the short term, we should get better at doing the things we know now. The 2015 MBRRACE (Mothers and Babies: Reducing Risk through Audits and Confidential Enquiries across the UK) confidential enquiry into term, singleton, normally formed, antepartum stillbirth indicated that half of the cases exhibited ‘major suboptimal care’ when compared against current guidelines.11 Ensuring high-quality implementation of current guidelines could have a major impact on stillbirth rates. Moreover, another key finding was that the inadequate review of losses was also common, and hence the lessons that could have been learned from losses were missed. It is essential that a thorough and effective review is performed after all losses, as this can identify weaknesses in systems within a hospital. MBRRACE is in the process of launching a standardised tool to help all units achieve this. It also follows that in a majority of stillbirths there was either no major flaw in the care or the event would not have been prevented even if care had been better. The main method for screening the general population for stillbirth is to measure the symphysis–fundal height (SFH) with a tape measure. But multiple lines of evidence implicate fetal growth restriction and abnormal placental function in the aetiology of many stillbirths, and there is very little high-quality evidence to suggest that the measurement of SFH, however it is performed, is useful as a screening test for abnormal placental function and associated fetal compromise.12 It seems likely that better tools for assessing placental function and the effects of adverse placental function on the fetus (including more sophisticated ultrasonic methods and/or biomarkers) could lead to the highly discriminatory screening tests that would be required. Hence, when addressing stillbirth rates, our resources should be focused on: (1) optimising high-quality care based on what we know now; (2) audit, and effective review of losses; (3) research addressing the clinical and economic effects of universal late pregnancy ultrasound; and (4) more basic research to identify better screening tools for future RCTs. None declared. Completed disclosure of interests form available to view online as supporting information. This paper was the sole work of GCSS. N/A. NIHR Cambridge Comprehensive Biomedical Research Centre. None. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.