Screened

Like millions of other students around the country, I was required to get a screening test for tuberculosis before starting school this month. I went to the corner family health clinic, got my injection, and came back a couple of days later. Unlike many other students, my test was positive. I couldn’t believe it—my hand was visibly shaking as we discussed the results. Would I be allowed in the school? Would I even be allowed near the other students? The nurse reading the test just laughed and told me not to worry; most medical students have been exposed to TB by their first year in the clinic, so I’ll just have a head start on the class. Just to be sure I didn’t have an active infection, however, she signed me up for a chest X-ray.

You’ve probably had one of these skin tests before. Often called a PPD test for purified protein derivative, the provider injects a droplet of fluid containing tuberculin bacteria proteins between the layers of your skin. When you return two days later, the clinic checks to see the degree of immune response your body has mounted against the foreign material. Usually, the spot is red for a few hours and then fades away: your body does not consider the proteins threatening. But if the spot remains raised, it means your immune system has responded more aggressively and in all likelihood has seen live tuberculosis bacilli before.

The chest X-ray showed no visible lesions in my lungs. For now, the best interpretation was that I had only a latent infection and was not contagious. Still, my lifetime risk of developing active TB was 10% unless I took a long course of anti-tuberculosis medications. “When could this have happened?” I wondered. A moment’s reflection provided some answers: I had spent all of last year working in a large hospital, and I had recently traveled in South East Asia, including some very rural areas. I could have easily been exposed in either of those places without realizing it, carrying a small colony of tuberculosis bacteria in my body unawares.

As a health minded medical student, I decided I must absolutely begin the anti-TB drug Isoniazid as soon as possible. It looked promising—finishing the course would reduce my lifetime risk of active TB from 10% to less than 1%. But the drug was not easy. I would have to take a pill every day for the next nine months. I would have some risk of neuropathy and anemia, but most certainly of liver damage, requiring a long hiatus from my enjoyment of craft beers (unthinkable!).

Still, it was necessary. Before starting the drug, however, they wanted to do one more test to make sure I really had TB, a test with much higher specificity. The “QuantiFERON-TB Gold” test required a small amount of my blood, which they mixed with TB proteins (similar to the ones injected into me earlier). If my white blood cells released lots of interferon (an immune attack signal) in response to the proteins, then I had indeed been exposed to TB.

Fortunately, they didn’t. A single phone call from the clinic made my year—I didn’t have TB! After I finished dancing in my living room, I was left with a list of questions: How could the two tests be different? What if I had taken the drug and had liver damage? Why even give PPD tests at all?

I had my brush with a false-positive, one of the reasons we don’t screen everyone for everything. Something non-intuitive and often surprising is that the accuracy of a test depends just as much on the population it is given to as on the characteristics of the test itself. For example, suppose a test with 99% sensitivity (ability to catch disease carriers) and 99% specificity (ability to identify people without the disease) is given to a population of 10,000 where 100 people have the disease (1% prevalence). Even with such an amazingly good test, only half of the people with a positive result would actually have the disease! This problem will always happen when a screening test is given to a population with a low prevalence. Test results don’t define your chance of having a disease, they adjust your risk based on the chances you came in with. This was very poorly understood by Illinois’ Gov. James Thompson in 1987, who enacted a law requiring HIV testing of all couples applying for a marriage license. Since heterosexual, monogamous couples were probably the lowest risk group besides cloistered clergy, even with a near perfect test there would still be an egregiously high number of false-positives. Indeed, there were, and most couples went out of state to marry until the law was repealed in 1989.

Even when good tests are available, expense is also a consideration. The PPD test is the mass screening test not because it is the most accurate, but because it is the cheapest. Since the main risk in the beginning is missing people who actually have the disease, the guidelines are set up so that nearly everyone who has the disease will have a positive test, plus a few people that don’t have the disease as well (a sensitive test). Only after the initial screening do we use the more expensive and more specific interferon test to knock out those false-positives who definitely don’t have the disease. Interestingly, it’s hard to tell what the actually accuracy of the different TB tests are, since there is no gold-standard (a supposed perfect test) with which to compare them to. A study used cross comparison statistics to try to find an answer, and you’ll see it’s nowhere close to our hypothetical example.

So, should we even continue giving TB screening tests to students, given that they are such a low-risk population in general? Once all the bad results are sorted out, the bottom line is that the TB test is useful because we can do something about the results—effective treatments are available. The clinical utility of a test is even more important than the accuracy when deciding on a screening program. There is a near perfect test available for Huntington’s disease, but 95% of at-risk people choose not to get it, since there is no treatment for the fatal illness. In the middle of all these questions is the debate over prostate cancer screening by PSA, in gray areas both for accuracy and clinical utility. 23andMe will scan my genome for markers associated with heart disease and diabetes, but could I really use the information that my lifetime risk is 9% instead of 7%?