The next two posts are about medical testing, something to which we are more and more subject as we get older. It seems rare these days to visit a doctor’s office and not come away with a handful of requisitions for some kind of medical test. This is especially true as we get older.
The array of information available from tests is light-years ahead of what was available when I started practicing medicine. That is generally beneficial for patients, but it has a significant downside. One of them is that doctors are much less trusting of their own clinical judgment, less adept at physical examination and at reading the signs and symptoms that a patient is presenting. I know I sound like an old curmudgeon lamenting a lost art, but physicians are much quicker to jump to testing, even though it is expensive and can be fallible.
To understand the limits of tests and to know how to interpret them, one has to start with an understanding of what “normal” and “abnormal” mean. We think of tests results as binary, as either normal or abnormal, abnormal indicating a problem and normal meaning all is well. Not so. Unfortunately, there is not a clean line between the two. All tests have an overlap between normal and abnormal, and, in fact, what is defined as “abnormal” is statistically determined, a compromise to include most of the people with disease and the fewest people without disease.
The people who actually have no disease and who turn up with an abnormal (positive) test are called false positives, and conversely, those with disease who show up as normal are false negatives. Virtually no test is perfect so as to eliminate this overlap, although the some tests are better than others. False positives are problematic because they designate a person who has no disease as having an abnormality, and this usually leads to more testing and sometimes invasive procedures like biopsies or even surgery to prove they are okay. False negatives are a problem because they are falsely reassuring.
An example would be a nuclear stress test in a person who had chest pain typical of cardiac disease. The test characteristics are such that 80% of the time, if a person has coronary disease as the cause of the chest pain, it will show up on a nuclear exercise test; 20% of the time, it won’t. So a certain percentage of people will be missed even with a test as sophisticated as an exercise test.
Now here is the tricky part. How well a test performs depends on the population being tested. Tests generally do better when there is a higher likelihood of disease. Suppose you are testing people who are very likely to have coronary disease, 80% likely based on the nature of their chest pain. And you perform a nuclear stress test that is 80% sensitive. It will pick up disease 80% of the time that it is present. The test is also estimated to be 95% specific; that means 95% of the normal patients will have a negative test. In this instance 94% of the positive tests will have coronary disease. Trust me with the math.
Now, using the same test and the same sensitivity and specificity, test a population where the disease is much less likely, let’s say 1 in 100 chance. Test 100 people and you still have a 80% chance of picking up the one with disease. However, 5% of those without disease will have a false positive test, meaning if the result is abnormal, it is 5 times as likely to be a false positive rather than a true positive. This is the big problem with scatter-shot testing, ordering a whole panel of tests when there isn’t anything specific that one is looking for. In people without symptoms, a significant number of tests will come back positive in normal people, leading to follow-up testing.
It is even a bigger problem with screening broad populations, for example screening all men over 50 for prostate cancer with a PSA. The chance of prostate cancer a random population of men is even less than 1 in 100, so the number of false positives goes up, leading to more testing and biopsies. It is the reason that many have recently questioned the advisability of mass screening with PSA tests, which leads to many prostate biopsies for a disease that progresses very slowly in most men. Testing has not been shown to reduce mortality.
I’m not arguing against screening when it is applied to high-risk people and specific age groups, but ordering a battery of tests as part of a physical examination may lead to more mischief than benefit. A pet peeve of mine is the selling of total body CT scans as a “screening” test for healthy people. One site advertises a total body scan for $425, and $799 for two people. It exposes people to considerable radiation and is bound to show “abnormalities” that are highly unlikely to be of significance.
So what are the takeaway points from all of this?
- There is no clean line between normal and abnormal for a test. Good tests will miss some people with disease (false negative) and be abnormal in some people without disease (false positive).
- Tests perform best when there is a reasonable likelihood that a person has a disease and worst when they are used indiscriminately where there is a low likelihood of disease. Consequently, screening tests are best applied to people who are relatively high risk.
- False positive tests have consequences. They lead to further testing, painful biopsies, or even erroneous attempts to treat an abnormality when there is none.
I’ll close with a case in point, my own. I was screened each year or two with a PSA, which was always normal until two years ago. At that time, the test number doubled, shooting me into the abnormal range. A repeat test showed the same thing. It was not very high, clearly in the overlap zone, but in one year it had doubled.
My choice was to get a biopsy or to wait and repeat the test six months later. If the PSA continued to rise, it would be stronger evidence that cancer might be present. A biopsy was recommended. I hesitated but went along with it. The biopsy was a painful procedure. I had no complications from it (excessive bleeding or infection), but the result was “borderline.” Some “atypical” cells showed up, which now prompted another biopsy six months later.
This one was no more pleasant than the first, but now the biopsy was normal. The PSA was repeated 6 months after the second biopsy. It was unequivocally normal, back to baseline. Of course, I was happy that all was normal, and I’ll never know why it was elevated in the first place, but I couldn’t help but think that if I had been a little more patient and waited a bit, the whole thing could have been avoided.
Al Martin, M.D., 75, is a former associate professor of medicine at University of California, San Francisco Medical Center and Chief Medical Officer of Blue Shield of California. He now writes a blog, Age With Spirit.