When receiving a test result from the new blood test for Down syndrome that says “positive,” do patients understand there may be a one-in-five chance of a false positive? Or even a one-in-two, depending on the mother’s age? Even genetic counselors admit this is difficult to grasp.
Yesterday, I featured an article from the New England Journal of Medicine raising concerns about Non-Invasive Prenatal Screening (NIPS), the new blood test for Down syndrome, being offered by Sequenom, Ariosa, Verinata, and Natera. The article cautioned against offering NIPS to every expectant mother. One reason noted in the article was that:
Arguably, PPV is more important than sensitivity and specificity to patients undergoing testing: it indicates the probability that a positive test result indicates a true fetal aneuploidy. Thus, PPV should be discussed in study reports and marketing materials but isn’t.
In a post on a blog for genetic counselors earlier this month, a genetic counselor further explained the importance of PPV and how it is not being understood with NIPS results.
Katie Stoll is a genetic counselor in Washington State. On the blog, The DNA Exchange, she examines the role of the incidence rate for Down syndrome and a test’s Positive Predictive Value (PPV).
The NIPS labs (and the media reporting on them) highlight their sensitivity and specificity levels as being greater than 99%–meaning their tests can identify greater than 99% of those pregnancies carrying a child with Down syndrome and rule out greater than 99% of those pregnancies not carrying a child with Down syndrome. But, unless some further calculations are done, a mother cannot understand how likely receiving a “positive” NIPS test actually means she is carrying a child with Down syndrome. As Stoll explains, this is because of the role of the condition’s incidence rate.
Incidence rate is how often a condition appears in the population. Down syndrome remains a rare condition, becoming rarer the younger the mother is.
Stoll takes as her example a population of 100,000 35-year old women who have an incidence rate of 1-in-250 carrying a child with Down syndrome. Therefore, of the 100,000 35-year old moms, 400 will be pregnant with a child with Down syndrome (100,000 X 1/250 = 400).
NIPS labs report a sensitivity rate of 99.5%, meaning 99.5% of those actually carrying a child with Down syndrome will be detected by NIPS. Therefore, of the 400 35-year old moms, 398 will receive a “positive” NIPS result (400 X 99.5% = 398). Note as well that 2 will receive a “negative” NIPS report–a false negative, since they are carrying a child with Down syndrome.
NIPS labs also report a 99.9% specificity rate–the percentage of those pregnancies not carrying a child with Down syndrome that will receive a negative NIPS report. In Stoll’s example, there are 99,600 moms not carrying a child with Down syndrome (100,000 moms – the 400 carrying a child with Down syndrome = 99,600). Of those 99,600 moms, 99,500 will receive a negative report (99,600 X 99.9% = 99,500). This then means 100 will receive a “positive” NIPS result (99,600 – 99,500 = 100)–making these 100 false positives.
So, in this example, there were 400 pregnancies actually carrying a child with Down syndrome. Of these, 398 would receive a positive NIPS result, but 100 false positives would also be reported, making for a total of 498 positive NIPS reports when only 400 pregnancies were actually carrying a child with Down syndrome. This means that a positive NIPS report means the mother has a one-in-five chance of a having a false positive (100 false positives / 498 = 20%, or 1-in-5). And, this false positive rate goes up the lower the incidence rate.
Take for example 100,000 expectant moms in their late 20′s. The incidence rate is about 1-in-1,000. This means there are only 100 moms actually carrying a child with Down syndrome (100 X 1,000 = 100,000). Of these, 99.5 would receive a positive NIPS result, given the 99.5% sensitivity rate. But, 100 of the moms who were not carrying a child would also receive a positive NIPS report (100,000 minus the 100 carrying a child with Down syndrome = 99,900 X 99.9% (specificity rate) = 100). So there are 100 true positives and 100 false positives reported. A “positive” NIPS result in this low risk population means only a 50% chance of actually carrying a child with Down syndrome (100 true positives out of 200 total positive reports = 50%).
Stoll’s post and the New England Journal of Medicine article make the same point: the accuracy of NIPS remains unknown because of the way the labs report both their research and their test results. The professional societies have called on the NIPS labs to standardize their reporting so the accuracy of each companies’ test can actually be determined. The NIPS labs have not done so, but as these articles, and hopefully this post demonstrates, a positive NIPS result is not a true positive.
Update: the subsequent posts in this series explain positive predictive value and its relation to maternal age, with graphs visually displaying the percentages. See these posts at this link and this link.