Introduction to non-invasive prenatal testing
One of the fascinating applications of modern genome sequencing is the testing of a fetus genome in pregnant women in a non-invasive manner. Millions of women have used this procedure worldwide, and it is doubtful that many would categorize it as a “genomic investigation”. Of course, non-invasive refers to the fetus, as this approach requires maternal blood to indirectly inspect the genome of the fetus (so it is still invasive to a pregnant woman, as most of us do not like needles!). The traditional gold-standard approach for testing for genomic anomalies is via the amniocentesis or chorionic villus sampling. While these approaches have long been established for diagnostic purposes, and provide a true fetal sample, they do carry a very small but real risk of miscarriage (0.5–1% for chorionic villus sampling, and 0.25-0.5% for amniocentesis).
Therefore you can imagine the excitement over the possibility of achieving similar results from just a maternal blood sample using next-generation sequencing. It is definitely an exciting topic for me, and it is exciting to finally introduce this information in my blog!
So how does it work? Our cells continuously die or are destroyed, and as they are broken down, their genetic material can be released into the blood stream. Once in the blood, the circulating DNA will be cleared up in matter of hours, but it is a process that occurs on a continuous basis. This allows for a constant supply of DNA to be in circulation, whether it is that of the mother, or of the developing baby. For the fetal DNA, it is on average about 10% of all circulating DNA.
This is why the analysis of fetal DNA isolated from maternal blood has been compared to looking for needles in a haystack. Powerful computational bioinformatic capabilities are required to accurately decipher that information. And on top of that, half of that fetal DNA was derived from the mother, reducing even further the fraction of fetal DNA that is unique and distinguishable against the maternal fraction. So technically, only about 5% of this will be unique, representing the inheritance from the father of the baby. Once all of that information is sequenced, it is a matter of measuring these fractions with appropriate statistical methods. In this battle of fractions, if there is the presence of aneuploidy, such as trisomy 21, the most common birth defect in the US, it will increase the fraction of DNA representing chromosome 21, and the abnormality can be detected.
It gets even more extraordinary than that! When detecting trisomies, an elevated fraction of DNA would be detected along the length of the entire chromosome. But what about just a fraction of a chromosome? It turns out that this technique is sensitive enough to detect changes in only portions of chromosomes too! Such sub-chromosomal microdeltions are the basis of different developmental defects in syndromes such as DiGeorge, Prader-Willi/Angelman, Cri-du-chat, Wolf-Hirschhorn, Jacobsen or Langer-Giedion. So it is a fairly robust and powerful tool, but I will leave this topic for another post! Data is out there, and many NIPT tests incorporate this information now as well.
So this sounds all great, but the latest stance of professional organizations such as the American College of Obstetricians and Gynecologists, the Society for Maternal-Fetal Medicine, and the National Society of Genetic Counselors state that while non-invasive prenatal screening is a safe and effective test in high-risk populations, they do not recommended it as a screening test due to the modest amount of published data evidence.
But that is no longer the case! Very large population studies have been released assessing both low-risk and high-risk populations, and they shed light on the situation. For clarity, the high-risk population is most commonly identified by advanced maternal age (> 35 years), but also by testing positive with conventional blood biomarkers screening for trisomies, producing an abnormal ultrasound, having a past family history of developmental delay, congenital malformations or genetic disorders, and/or past pregnancy outcomes, including recurrent pregnancy loss which could be indicative of chromosomal alterations. But maternal age is obviously in this day and age the most common reason.
In the past, the studies testing the efficacy of non-invasive prenatal tests were rather small in scale. One of the outcomes was that in low-risk populations there was a high rate of false positives, and hence low positive predictive value. False positive is when a test states there is a certain outcome which is actually not there. There can be unusual circumstances that can obscure results. One of them is the fact that it is actually not fetal DNA that is being measured directly, but rather what is believed to be placental DNA. Since both sources came from the same fertilized female egg, it is used as a proxy for accurate fetus genetic material. But in rare instances, if alterations in the genetic code occur so early in the development, it can lead to populations of cells with differing genetic material. This is referred to as mosaicism, and has clearly been shown that placental mosaicism can lead to wrong test results.
Have you ever picked up a fruit or vegetable that had a distinct patch of different color than the rest of its skin? Well, imagine this different pigmentation is caused by a mutation that resulted in different genetic background and a different fruit color. How much of the skin area results in a different color would depend upon how early in the fruit development the mutation has occurred. And such spontaneous mutation events can happen in any organism, at any stage of existence. If it occurs early enough in the development, including during the formation of the placenta, you will end up with two different genetic make ups.
A false negative is when a test states there is no abnormality detected when it is actually present. This is the tricky one because it can lead to a false sense of security and not be verified, whereas when a false positive outcome is provided, the result can be verified with a different test.
Most of the time, the measure of a test is provided in terms of sensitivity and specificity (which measure the avoidance of false negatives and false positives, respectively), but these are based on whole population studies. So the positive predictive value of a test is the metric to truly look at as it provides the information on how often a positive result can be expected to truly be a positive result.
This measure not only incorporates the sensitivity and specificity of a test, but is also highly dependent on the prevalence of measured conditions (meaning, how often a given condition is actually observed in a population). So for a non-invasive prenatal test in a low-risk population, where the incidence of aneuploidy is less common than in a high-risk population, the positive predictive value could be lower than for the high-risk population. This is the reason why traditionally the test has been primarily offered to high-risk pregnancies.
In fact, NIPT screening has attracted some controversy over the issue of its positive predictive value, or rather lack of proper education around this topic. Many recipients, and even some of the providers, do not fully understand how the test works, quoting the sensitivity of the test as a probability of detection instead of the positive predictive value. And as you will see below, that can be very misleading.
This might be a weird concept to wrap your head around at first. But basically, if the test parameters stay the same, what it means is that the rarer a condition we are trying to detect, the more likely it is that we will have false positives (we will falsely claim that a disease is present when it is not). Why? Because we are so keen to find the disease when it is really there, that we are willing to make the mistake of saying it is there even if it is not, as opposed to the mistake of not calling it when it is there (the biggest worry). So while the real disease might have been detected by the test, and no false claims are made that disease is absent when it is present, the predictive value of the test still drops just because of all the false claims of disease being present when it is not.
Like I mentioned, it is a weird concept to wrap your head around! So let’s toss in a visual to help explain this concept. The prevalence of trisomy 21 is much higher than that of trisomy 13 for any age group. So assuming that 100 000 high-risk women were screened with a test that exhibited 99% sensitivity and 99.9% specificity, you can see that even though all affected cases might be correctly detected for trisomy 13, all the wrong calls (red circles) in the unaffected population negatively impact the test positive predictive value.
So let’s look at some of these values based on couple examples of large population studies. One such study recently assessed the results for nearly 150 000 Chinese pregnant women with an average maternal age of 31 years, with 95% of the pregnancies assessed in the second trimester.
The women were assessed for trisomies of chromosomes 21 (Down syndrome), 18 (Edwards syndrome) and 13 (Patau syndrome). The measured incidence of these events was 0.64%, 0.15% and 0.02%. Overall sensitivity was 99% and specificity was 99.9%, while the positive predictive value was 85%, so it confirmed the disorder in 85% of the cases. So you can appreciate how misleading it can be to advertise a test by quoting a 99% rate instead of 85% as people automatically tend to interpret these numbers as test accuracy levels.
But going back to the meaning of the positive predictive value, luckily, the 85% value is due to false positives and not false negatives. It is the difference between getting fake news (which you can verify), versus the omission of really important news (which you are unlikely to double-check). In total, just over 900 cases were correctly identified, just over 150 false positives were called (incorrectly informed of the potential trisomy issue), and only 9 false negatives were presented (where the condition did exist but was not identified). This means that the test is very good at also telling when the condition is not present (99.99% of the time in this case). This is actually a great number to brag about, as after all, this is the information that women are most keen to receive!
As you can see based on the incidence rate of different trisomies in the population, the positive predictive value varied for different cases of trisomy. It was 92.2% for Down syndrome, 76.6% for trisomy 18 and only 32.8% for trisomy 13. In the case of trisomy 18, while no false negative was observed, 22 correct cases were identified, but 45 false positives were also called where the condition thought to be present was not there (as confirmed by secondary testing and postnatal assessment). A similar number of false positives were called for each of the trisomies, so you can see how the prevalence of a condition can impact the positive predictive value. What is great is that the authors also looked at the numbers for Down syndrome in high-risk versus low-risk women, and showed positive predictive values of 94.1% and 81.4% respectively. Due to this low predictive value in low-risk women, the non-invasive prenatal test is not considered a diagnostic test, but is suggested to be an appropriate screening test in a general population, including for low-risk women.
But nevertheless, organizations such as the Society for Maternal-Fetal Medicine still do not recommend the use of such tests for low-risk women. Their latest update on the topic has in fact assessed this and other studies, but justify their position due to the “commercial interests” of laboratories involved, and a “limited transparency of many details”, although I would not be able to make that last statement with regard to the discussed study above.
Whether used for the high-risk population only or not, the bottom line is that non-invasive prenatal screening has become an integral part of obstetric clinical practice and is bound to witness increased use in the future, especially as its cost continues to decline. This is great news that another powerful tool has been added to the arsenal of prenatal testing for expecting mothers.
And guess what? You can even supplement your regular NIPT screening with additional information by sequencing the fetal DNA and looking at what informative mutations are discovered. The world of measured possibilities is certainly rapidly expanding! But this will also be a topic for another blog post. In the meantime, if you are interested, contact Merogenomics, and we will let you know where, how, why and for how much. For the NIPT test, talk to your doctor, although don't be surprised if you do not hear positive predictive value mentioned. Luckily you have this post! Happy baby genomics. ;)
This article has been produced by Merogenomics Inc. and edited by Kerri Bryant. Reproduction and reuse of any portion of this content requires Merogenomics Inc. permission and source acknowledgment. It is your responsibility to obtain additional permissions from the third party owners that might be cited by Merogenomics Inc. Merogenomics Inc. disclaims any responsibility for any use you make of content owned by third parties without their permission.
Products and Services Promoted by Merogenomics Inc.