Cell‐free DNA testing: inadequate implementation of an outstanding technique

In recent years, a powerful molecular technique based on fetal DNA analysis has been introduced in prenatal diagnosis; the analysis of fetal cells in the maternal circulation, a kind of ‘holy grail’, pursued for decades, has shifted to the detection of cell-free fetal DNA (cfDNA) in maternal blood1. Briefly, fragments of placental DNA are analyzed in order to establish the number of specific fetal chromosomes. The main aim of cfDNA testing is to reduce the number of invasive procedures (amniocentesis or chorionic villi sampling (CVS)), increasing the safety of genetic analysis. In parallel with the introduction of cfDNA, to increase the broadness of genetic analysis, another new molecular technique based on fetal DNA analysis has been introduced: genomic microarray analysis. The main objective of microarrays is to diagnose smaller genomic imbalances not detectable by standard karyotyping. Despite the established contribution of both techniques in prenatal diagnosis, their introduction into routine practice differed considerably. cfDNA testing for aneuploidies was first described in 20082, 3 and began to be implemented in clinical practice in late 20114-7, only 3 years later, while genomic microarray analysis, first described much earlier, in 1997, was applied to prenatal diagnosis in 2005, and is still underused in the context of fetal malformations. Thus, the implementation of cfDNA received an unprecedented welcome by both the scientific community and public opinion. However, despite the large-scale promotion of cfDNA, concerns have been raised regarding the complete shift in the research paradigm and several drawbacks have not yet been addressed sufficiently. The marketing of cfDNA has been based on five misconceptions. The first concerns the aim of prenatal screening. For several reasons, people identify Down syndrome with neurocognitive impairment, and most women are more worried about neurocognitive impairment in general than they are about Down syndrome itself. Down syndrome is certainly the most common cause of neurocognitive impairment, and its phenotype is recognized universally. This misconception does not apply solely to cfDNA, but encompasses all Down-syndrome screening methods, which were based initially on advanced maternal age and later on biochemical and ultrasound screening. Yet, historically, basic prenatal screening for birth defects placed emphasis on common disorders such as Down syndrome and, among major structural anomalies, neural tube defects. In other words, the first common misconception is the identification of Down syndrome with neurocognitive impairment. The second misconception derives from the confusion between screening and diagnostic methods. cfDNA was marketed initially as ‘non-invasive prenatal diagnosis’ and it has been advertized repeatedly in the press as the method to replace amniocentesis. Thus, it has been reported that 7% of screen positives by cfDNA opt for termination of pregnancy without proper confirmation7, 8. Some referring physicians may even treat the test as diagnostic, either not taking into account that the positive predictive value (PPV) may vary from 30% to over 90%, or in response to patients who are unwilling to wait long for results. cfDNA indeed has higher detection rates and lower false-positive rates (FPR) than previous screening methods, such as the combined test (99% vs 90% detection rates for Down syndrome for a 0.09% vs 4% FPR, respectively9), but it still requires confirmation by amniocentesis or CVS. It is also very important to bear in mind that, although in low-risk populations sensitivity and specificity are similar to those in high-risk populations, the PPV is much lower, at < 50%6 compared with > 90%10. Thus, the second misconception is that cfDNA is a diagnostic method, replacing invasive procedures, instead of an advanced screening method. The third misconception involves the term ‘non-invasive prenatal testing (NIPT)’, the name having been selected to describe the non-invasiveness of the new technique, rather than its intrinsic characteristics. This term is redundant, since all screening methods in prenatal diagnosis are non-invasive from the fetal point of view. The Merriam–Webster medical dictionary11 defines non-invasive as: ‘done without cutting the body or putting something into the body’. On this basis, blood sampling for cfDNA testing should be considered a maternally invasive procedure, as it entails sticking a needle into the woman's arm. Differentiating between fetal and maternal invasiveness is clinically relevant, given that the former entails some degree of fetal loss, yet maternal blood sampling is definitively an invasive method. Nevertheless, the term ‘non-invasive testing’ has been very useful for marketing purposes, since it is easily understood, and any non-invasive procedure is expected to be well-accepted by both partners. It appears that many scientific societies and researchers are moving to ‘cfDNA testing’ as a more accurate term for this method9, 12. Thus, the third misconception involves considering cfDNA testing as being the only non-invasive screening method and maternal blood sampling as non-invasive. The fourth misconception concerns a failed result (known as a ‘no call’, ‘cancellation’ or ‘failed quality-control metric’), given that in clinical practice this is a sort of positive screening result. Strictly speaking, the FPR (i.e. percentage of unaffected pregnancies with a screen-positive result) is directly proportional to the number of aneuploidies tested for. According to a recent meta-analysis, the FPR is about 0.09% if only trisomy 21 is assessed, increasing to 0.72% if trisomies 21, 18 and 13 and sex aneuploidies are tested9. This rate increases substantially if cfDNA testing is expanded to detect common microdeletions13, given that it has been shown recently that there is a 1% excess FPR if only as few as five microdeletion syndromes are studied14. According to this large meta-analysis including more than 40 000 samples from 31 studies in singleton pregnancies, an extra 5.1% should be added to these values if we consider as screen-positive those pregnancies with a failed result, given their reported 4% risk of aneuploidy9, 15. Laboratory failure to obtain a result may be due to low fetal fraction (< 4%), typically seen in trisomy 18 and triploidy, or to assay failure, each accounting for half of the cases9. Not measuring the fetal fraction raises the issue of lack of quality assurance rules and has led to erroneous identification of samples from non-pregnant women as those of a normal female fetus16. Some have suggested, in cases of failed results, to rely on the ultrasound findings, as abnormalities are expected in trisomy 18 and triploidy17, however, this does not take into account that most of those fetuses do not show evidence of anomalies before 18 weeks, and some even later in pregnancy12, 17. In addition to laboratory failure, 4.3% of samples are inadequate9. It is crucial for fetal medicine experts to provide parents-to-be with accurate information on both ‘invasive’ and ‘non-invasive’ techniques. Historically, a 1% increased risk of fetal loss has been quoted for invasive procedures. This risk was derived from a 1986 publication of a single randomized trial completed in May 198418 (the month in which Mark Zuckerberg was born, and the Soviet Union announced the boycott of the 1984 Los Angeles Summer Olympics). The trial compared fetal loss in pregnant women undergoing amniocentesis by means of an 18-G needle with that in controls, the increased risk value of 1% being obtained from the finding of a significant difference between the 1.7% fetal-loss rate in cases and the 0.7% rate in controls. Recent data demonstrate that amniocentesis and CVS no longer entail a 1% risk of fetal loss. In 2006, a prospective sub-study of the FASTER Trial19 concluded that the procedure-related fetal-loss rate up to 24 weeks after mid-trimester amniocentesis was 0.06%. Specifically, no significant difference in loss rates was found between women undergoing amniocentesis and controls (1% vs 0.94%). The results of this study received more criticism than support in the several commentary letters that followed the initial publication. Since then, most of the controlled studies addressing this issue failed to confirm the 1% fetal-loss risk ‘dogma’, after both amniocentesis and CVS. Thus, in 2008, the Washington University group in St. Louis published similar results, based on their experience of 11 746 amniocentesis and 5243 CVS procedures. They concluded that the 0.13% fetal-loss rate attributable to amniocentesis and the 0.7% attributable to CVS were not significantly different from those observed in pregnant women with no procedure20, 21. Interestingly, a study from Nicolaides' group showed that most of the fetal losses after CVS could be predicted by maternal and pregnancy characteristics22. A prediction model was developed among 33 856 singleton pregnancies scanned at 11–13 weeks. The risk was increased with certain maternal factors (greater age and weight, African racial origin, previous miscarriage or stillbirth, pre-existing diabetes mellitus and conception with ovulation induction) and with abnormal first-trimester screening markers (low pregnancy-associated plasma protein-A levels, increased nuchal translucency thickness and reversed ductus venosus A-wave). Subsequently, this model was applied to a group of 2396 patients undergoing CVS and the predicted number of miscarriages was 45, not significantly different from the observed 44. In line with those previous series19-22, a recent study from Northwestern University demonstrated no increased risk for unintended fetal loss in women undergoing CVS compared with a suitable control group of women with at-risk pregnancies undergoing cfDNA (0.92% vs 1.17%)23. The recently published meta-analysis by Akolekar et al.24 confirms that the procedure-related risks of miscarriage following amniocentesis and CVS are much lower than the traditionally quoted values. They found the estimated risks of miscarriage prior to 24 weeks in women undergoing amniocentesis and CVS to be 0.81% and 2.18%, respectively, while the background rates of miscarriage in controls not undergoing any procedure were 0.67% for the amniocentesis group and 1.79% for the CVS group. The weighted pooled procedure-related risks of miscarriage for amniocentesis and CVS were 0.11% (95% CI, –0.04 to 0.26%) and 0.22% (95% CI, –0.71 to 1.16%), respectively, neither being statistically significant24. Women should be provided with accurate and up-to-date information on both invasive and non-invasive prenatal diagnostic testing so that they can make evidence-based choices. A defensive medical approach suggesting disproportionately high fetal-loss rates for invasive procedures should be avoided in clinical practice, although it is true that there are no invasive procedures with zero risk. It is worth wondering why, nowadays, there is so much anxiety regarding invasive procedures, given that only 10–20 years ago it was considered perfectly acceptable to perform CVS and amniocentesis for no other indication than advanced maternal age. There could be several reasons: since the introduction of cfDNA testing, the press has been emphasizing that expectant mothers no longer have to face the inconvenience of an invasive procedure and the possibility of procedure-related pregnancy loss; doctors, under these circumstances, adopt a defensive medical strategy, while financial pressure in the private sector is somehow challenging healthcare provision. However, emphasis should be given to the appropriate training of fetal medicine experts in both invasive procedures and counseling skills, to enable parents to opt for optimal individualized choices. It must also be borne in mind that the outdated 1% fetal-loss rate is typically quoted in cfDNA cost-effectiveness analysis; thus the resulting theoretical number of losses avoided by cfDNA is in fact considerably inflated25. Thus the fifth misconception regarding cfDNA is associated with outdated information regarding the impact of invasive diagnostic procedures. After the pioneering detection of trisomy 21 in maternal plasma by Dr Lo's group in 20082, cfDNA was introduced into clinical practice within 3 years, in 2011. The initial implementation process and validation studies were undertaken by the industrial research labs of BGI in China and of Sequenom in the USA4. This process attracted enormous resources in cfDNA research, allowing rapid progress in the field. In times of great financial recession, when research funds are declining worldwide, this industrial contribution was welcomed. However, some concerns regarding research methodological issues were expressed recently by Benn et al.26, given that ‘generally, the samples were not tested blindly, and the outcome and ascertainment criteria were not fully described’. Previous screening methods were described and implemented by academic centers in the UK and USA, with prominent professors in the lead, namely Professors Haddow, Wald, Cuckle and Nicolaides. In contrast, industry has been leading, funding and supervising the implementation process and most of the validation studies for this new screening technology. The implementation process has been unusually fast, but financial issues could compromise service provision26. Another concern involves the commercial promotion of this new technological advance. Press releases worldwide declared that the time has come to abandon amniocentesis (CVS is still unknown to journalists) due to the introduction of cfDNA testing. A recent study from the UK analyzed press articles for cfDNA in the 10 most circulated print/digital news media, and revealed that only one third of them were considered balanced regarding benefits and limitations27. It seems, therefore, that the marketing of cfDNA testing directly to the public was problematic. Furthermore, the large-scale promotion and the great demand among pregnant women did not allow for a structured educational program to be addressed to doctors and midwives. Fortunately, scientific societies have played a major role in establishing the limitations of cfDNA. It has been stressed that cfDNA is a screening method, always requiring confirmation by an invasive procedure in case of positive results, despite sensitivity and specificity of above 99%. Societies such as the International Society for Prenatal Diagnosis suggested that the term ‘non-invasive prenatal testing (NIPT)’ would be more appropriate than the initial term of ‘non-invasive prenatal diagnosis (NIPD)’28. The term ‘non-invasive DNA testing (NIDT)’ has been suggested as an alternative option. Scientific societies have also raised concerns regarding the application of cfDNA in low-risk populations, since almost all of the validation studies have been conducted in high-risk populations28-30. However, recently, prominent sessions in conferences organized by scientific societies have been hosting industrial research labs to support further development of this novel accomplishment. Obviously, a serious potential drawback of industry-led research involves the difficulty of achieving collaborative studies between different companies, due to the high competition between them and the current litigation over intellectual property. Another point to highlight is the paucity of population-based data on cfDNA testing and clinical and pregnancy outcomes, due to the fragmented and commercial nature of the providers. Hopefully, in the future, uncertain issues, such as those surrounding fetal mosaicism, will benefit from knowledge sharing. Thus, implementation of cfDNA testing represents a complete shift in the research paradigm: unlike previous first- and second-trimester screening programs, cfDNA testing is based on industrial funds to enhance its development, allowing for its fast commercial promotion. There are two barriers to expansion of cfDNA accessibility to the whole pregnant population: high cost and centralization of testing to a limited number of labs. It is well established that any effective screening method requires availability and affordable cost, conditions which are, for the moment, unfulfilled by cfDNA testing. The cost of a single cfDNA test is around $1000 or 800 €, while the cost of the combined test can be less than 50 € in countries in which the first-trimester scan has long been implemented (e.g. Germany, France, Italy and Spain). It is certainly unrealistic to expect public-health services to cover this fee for every pregnant woman, particularly, for example, in Southern European countries, where the annual public health expenditure per capita is between 1000–2000 €. While private insurance in the USA usually covers part of the cost, there is no consistent policy in Europe. Thus, in many countries, the cost of cfDNA testing is borne by the pregnant woman, which could compromise equity in healthcare. Further commissions rewarded by labs to private practitioners in some countries could make the whole story more complicated. A cost-effectiveness analysis carried out by Cuckle et al.31 in 2013 showed that the marginal cost of Down syndrome detection, defined as the cost of the incremental detection by universal cfDNA screening, compared with the first-trimester combined test, is as much as 1.64–7.77 million US$31. This cost can be lowered to 0.44–1.58 million US$ if the test is offered contingently to 10–20% of the pregnant population. Another cost-effectiveness study from Canada supports this contingency model, showing it to improve the current performance of first-trimester combined screening by increasing the number of trisomy 21 cases detected and reducing the number of invasive procedures, at a modestly increased cost ($22 000 for each additional case diagnosed prenatally)25. A second potential barrier to wider access is the centralization of testing to just five laboratory headquarters: four in the USA (Sequenom, Ariosa, Verinata, Natera) and one in China (BGI). This global phenomenon has transformed local genetics labs into DNA forwarding companies, sending samples overseas. These features make the current cfDNA testing model an unsustainable screening method. The launch of some academic labs in Belgium and The Netherlands is a promising sign of a new decentralized and decommercialized model32. The Dutch experience is a particularly interesting project, as initially it promotes a nationwide education program, addressed to midwives and practitioners, before the introduction of a sample collection process. In addition, the UK RAPIDR study33 is an example of publically-funded research that implements several cfDNA methodologies in its open-source bioinformatics program, making this technology more accessible. On the introduction of cfDNA testing into prenatal care, its large-scale promotion was facilitated by the emergence of several misconceptions regarding its aim and statistical and technical descriptions. However, its rapid implementation, resulting in discordant diffusion in relation to other established DNA prenatal diagnostic techniques, should prompt a reserved attitude in the scientific community. The anticipated widespread use of microarrays, due to the 1% increase in detection of pathogenic findings, never happened34, 35. Industry is welcome to contribute to the development and implementation of new techniques, but close surveillance by well-established scientific institutions is mandatory. Clinicians should promptly acknowledge both the indications and limitations of screening techniques and consider only evidence-based approaches in their routine clinical practice and pretest counseling. The introduction of non-invasive genomic analysis technology, combining cfDNA and microarrays into a non-invasive microarray, will no doubt provide in the near future a powerful tool in prenatal diagnostics.