False positives in prenatal tests: Study reveals reason for flaws

For expecting parents, pregnancy can be a time of delicious anticipation. Questions pop up, to be answered only with time. Boy or girl? When will she arrive? Will he have my eyes?

And then there’s the one keeping many parents up at night: Is my baby healthy?

Recently commercialized prenatal screening tests that analyze fetal DNA in the mother’s blood promise to answer that question as early as 10 weeks of pregnancy.

But a new study from researchers at Fred Hutchinson Cancer Research Center and the University of Washington published this month in the New England Journal of Medicine has found that quirks in a mom’s DNA can alter those tests results, leading parents — and even their doctors — to believe their babies have major disorders when they are, in fact, healthy.

Importantly, the test manufacturers could immediately alter their analysis to rule out this particular type of false positive, said Dr. Hilary Gammill, a Fred Hutch and UW obstetrics researcher who ran the study along with UW geneticist Dr. Jay Shendure. 

Those false positives are a big issue, Gammill said, because of the way the tests are marketed. Gammill said her OB/Gyn colleagues saw this first-hand with women visiting the UW obstetrics clinic for prenatal care who had diagnostic tests done that definitely verified the existence of a problem, or confirmed the fetus was normal.

“Because the test performance accuracy was billed as so high, when there was a false positive, the impact was durable through the whole pregnancy,” Gammill said. “It was hard for people to accept that things were really normal even when they were.”

Built on a faulty assumption

Because of the emotional toll of false positive results, Shendure, Gammill and their colleagues decided to take a closer look. Nobody knew how those false positives were arising, but the researchers had a few ideas.

“We tried to figure out if there were some that could be fixed,” Gammill said.

The privately manufactured tests, which include Sequenom Inc.’s MaterniT21 and Illumina’s verifi, flag babies likely to have major chromosomal defects. Such disorders include trisomies, conditions where a baby carries an extra copy of a chromosome and which can be fatal or severely debilitating.

The tests all look at cell-free fetal DNA — pieces of baby’s DNA that pass through the placenta and circulate in the mother’s blood. But moms have cell-free DNA in their blood too, mixed with and often indistinguishable from their babies’ (after all, half of baby’s DNA is identical to hers).

To identify too many or too few chromosomes, the tests ask whether there’s the right proportion of genetic material coming from the chromosomes most likely to be altered in a genetic condition. Trisomies of chromosomes 13, 18 and 21 are by far the most common (other trisomies nearly always result in very early miscarriage).

That analysis is predicated on a simple but flawed assumption, the authors wrote — that every woman has the same amount of DNA. In fact, some people can have more than the “normal” amount of DNA and never know it, due to harmless duplications of pieces of their chromosomes.

In their study, Shendure, Gammill and their colleagues looked at four women who’d received false positive results of a trisomy for chromosome 18. They found that two of these women indeed had extra DNA on one of their own copies of the chromosome. One of the women’s duplications was so large that it effectively increased her probability of getting a false positive on the test to 50 percent, the authors wrote.

All four women went on to have healthy babies. While the researchers still don’t know exactly how often these particular chromosome duplications influence test results, they used computational models to show that the phenomenon might underlie a substantial fraction of these tests’ false positives. The investigators hope their results will spur researchers and the test manufacturers to study larger numbers of women to answer that question.

Unclear results

Although these blood tests are meant for screening purposes only, meaning they cannot definitively diagnose a disease or disorder but rather only point to the likelihood the condition exists, the test manufacturers market the products directly to patients and often don’t clearly explain their drawbacks, touting instead the most impressive sounding of the tests’ statistics.

And most patients don’t understand those statistics, or the difference between screening and diagnostic tests, such as amniocentesis and chorionic villus sampling, which can definitively reveal a condition. New tests are coming on the market so quickly that even doctors and genetic counselors are challenged to keep up with all of the changes.

“Patients may say, ‘I chose to do this test because of the test performance and now I’m faced with a positive, and where does that leave me?’” Gammill said. “It can be just an incredible amount of stress on those patients.”

Gammill doesn’t know any women who terminated their pregnancies based solely on the results from prenatal screens, nor has she heard of any from other doctors she knows. Although this is likely a rare occurrence since experts always recommend a true diagnostic test to follow up a positive screening result, an investigation by the New England Center for Investigative Reporting stated that three women at Stanford University had aborted healthy pregnancies after receiving high-risk results from the screens.

It’s true that this new generation of prenatal screening tests is much more accurate than the tests’ predecessors, Gammill said. Another study published alongside Gammill and Shendure’s confirms that accuracy.

But the marketing is murky. For example, although the previous incarnation of screening tests were also noninvasive, Sequenom’s patient materials compare MaterniT21 PLUS to amniocentesis, which is a diagnostic procedure, not a screening test. Despite their comparison, women who test positive on MaterniT21 still need to undergo amniocentesis or a similar invasive procedure to confirm their result.

“If you look at the websites of the companies, they’re not trying to say they’re diagnostic, but the language is really all about the accuracy,” Gammill said. “So there’s this disconnect.”

The verifi website advertises that the test has near or above 99 percent sensitivity and specificity. Those numbers don’t tell you if you get a positive result, how likely it is to be true, Gammill said.

That particular statistic is not nearly as black and white, especially for women under 35, who are at low risk of pregnancies with chromosomal defects. Two studies looking at positive results of the tests in these women found that the “positive predictive value,” that is, the likelihood of a positive result being true, ranged from 40 to 75 percent.

So depending on the type of disorder and her age, a woman facing a positive from a screen that promises “accurate results, the first time” could have up to a 60 percent chance of a healthy baby. But women may not get that full picture, Gammill said.

‘This should be irrelevant really quickly’

She believes their study will alleviate some of that confusion. Follow-up studies on larger numbers of women to verify the phenomenon could be done rapidly, because the companies have all the relevant data in hand, she said.

The researchers haven’t been in touch with the test manufacturers, but Gammill said they hope their study will spark that conversation. And they’re optimistic about the results. The companies would only have to change the way they analyze their data to eliminate this type of false positive, she said. It’s a relatively easy change.

“Their goal is to provide an accurate test,” Gammill said. “We feel like this paper should be irrelevant really quickly … This is amenable to being fixed.”

The study’s lead authors were Matthew Snyder, UW genome sciences graduate student, and Dr. LaVone Simmons, former fellow in Maternal-Fetal Medicine at UW.