A study of patients at high risk for lung cancer has been announced to the lay press. It showed that screening current and former heavy smokers with low-dose computed tomography (CT-scans) resulted in 20% fewer deaths from the disease compared with a standard chest X-ray. Before you visit your local imaging center after buying a carton of cigarettes stop and catch your breath. There’s not likely to be much here.
To begin with the data have not been published in any medical journal. Thus they have not been reviewed by disinterested workers in this field. Next consider that most medical screening tests don’t pass rigid scrutiny. This doesn’t mean they aren’t done, rather many are performed on the basis of weak evidence about their effectiveness, cost, and side effects. The foremost problem with screening tests is that of false positives. Without going into much about statistical methods, lets define sensitivity, specificity, and positive predictive value.
The sensitivity of a test is that percentage of patients with the disease who will test positive using the screening test employed. If a test has a 95% sensitivity (virtually no screening test has a sensitivity that high) then 95% of the patients with the disease will test positive. Assume the specificity for the test is also 95%. Specificity is the percent of patients who don’t have the disease who will test negative. In this scenario 5% of patients with the disease will test negative; but 5% who don’t have the disease will test positive. And that’s the rub. Suppose you’re testing a million patients for a disease that 1% have – a 1% prevalence any particular disease is quite high. Thus 10,000 patients have the disease – 9,500 of them will test positive; 5% of those screened who don’t have the disease will also test positive – ie, there will be 50,000 false positives. The positive predictive value of a test is the number of true positives divided by the number of both true and false positives (multiply by 100 to get percent). Thus in the above example 9,500 divided by 59,500 is about 16%
Sticking with this example, if you test positive for the disease in question you have less than a 20% chance of actually having the disease even though the test used had a sensitivity and specificity of 95%. If the sensitivity and specificity of the test is 90%, which is still higher than almost medical screening tests, then the positive predictive value is less than 10%.
Some epidemiologists have estimated that the positive predictive value of a positive mammogram is not much more than 7%. This is because there are so many false positive mammograms. If doctors (who likely slept through their statistics course if they even bothered to go) and their patients understood the low positive predictive power of mammography I suspect the enthusiasm for the test would be considerably less than it now is. Nobody pays much attention to epidemiologists. They are to medicine what accountants are to hard driving businessmen. Whenever they try to explain reality to oncologists and radiologists they are shouted down and left to sulk in a corner. But the questions they raise will not go away even if they’re ostracized.
Is there an effective treatment for the disease you may have and is this treatment more effective if applied early rather than later in its course? Is the disease you’re screening for enough of a health problem to justify the effort of screening for it? Are the costs that will result from screening low enough to warrant screening?
So you’ve tested positive for a disease that you have a 10 to 20 % chance of actually having, do we stop or do we dig (sometimes literally) deeper at the cost of time, money. pain, and suffering to discover whether your afflicted or selected? The answer is obvious. But even if you’re willing to accept all these negative consequences we still have several hurdles ahead of us.
Lung cancer is the leading cause of cancer death in the US. The major cause of this disease is smoking. How many smokers should we screen? What predictive power will we accept? How much money are we willing to spend on such an effort? How many lives will be saved or at least lengthened? All these questions and many more require resolution before we add another burden to our already overburdened medical system.
My guess is that CT screening smokers will not prove medically or fiscally useful. But that doesn’t mean we won’t do it. I suspect that it will start even before the basic questions outlined above are answered. In fact I’ll bet it’s already started.