by Corinne Keet
Did you think my last blog post about the risks of the direct-to-consumer diagnostic testing movement as exemplified by Theranos oversold things? Was too abstract? Well, this story is for you.
Apparently, people are having abortions after getting positive prenatal genetic screening tests without confirmatory testing because they and their doctors do not understand how to interpret diagnostic testing. In particular, they do not understand the difference between specificity and positive predictive value. As I mentioned in my previous post specificity is a measure of how often people without a disease will test negative for it. But, if you’re the person with a positive test, what you really care about is positive predictive value, which is the likelihood that if you test positive you really have the disease. The problem is that positive predictive values are dependent on how likely you are to have the disease before taking the test. Unless your test is virtually perfect, when screening for a very rare disease, most positive tests are going to be false positives.
Let’s take an example with some math. Let’s say a test is 99% specific (“accurate”) like the one mentioned in the article. Let’s also say that the prevalence of the disease is 1/2,500 (estimated to be the prevalence of Trisomy 18). That means that for every 10,000 people tested, 4 will have the disease. To make things simple, we’ll assume that all 4 people who have the disease test positive. Because the test is 99% specific, 99% of the rest of the people without the disease will test negative (good for them), but 1% will test positive, which in this case is about 100 people (10000-4 = 9996, 1%*9996 = ~100). Let that sink in: 100 people who do NOT have the disease will have a positive test.
We can also calculate the positive predictive value, which is the likelihood that a test is a true positive. Since 4 will have a positive test and actually have the disease and 100 will have a positive test but not have the disease, the positive predictive value is: 4/(100+4) = ~4%, so that 96% of the positive tests will be false positives. Even if the test is 99.9% specific, the positive predictive value is still only 4/(10+4)=29%, and 71% will be false positives.
Bottom line: we need to be really careful about interpretation of screening tests and educate patients, health care workers and the public about their limitations. No one should be making a decision about something like terminating a pregnancy based on misunderstood screening tests.