If you read the previous blog from the screening iconoclast and know a little about screening, you might be thinking that the thing to do is focus on the positive predictive value rather than the sensitivity of your test.
The positive predictive value or PPV is the proportion of people who test positive on the test who have the disease you are looking for. One of the reasons it is not widely used in research papers describing new medical tests is that the PPV depends on how common the disease is in the population that you are using the test in. To understand that, consider a test that is no better than tossing a coin – it actually tells you nothing about whether someone had disease. Still, you decide to use it in a group of patients who have suspected cancer. If 40% of those with suspected cancer actually have that cancer then the PPV will be 40%. If on the other hand you decide to use it for a medical check-up in people with no signs of cancer and in whom only 1 in 2500 will have undiagnosed cancer, then the PPV would be 0.04%. The PPV tells us as much about the target group as it does about the test.
If, however, it is agreed that we want to use a test in a screening setting where we know how common it is to have cancer, then the PPV may be useful. In such a setting it is useful to think about how the probability of having cancer changes as a result of knowing the test result. Suppose, in an early diagnosis setting, someone presents with symptoms that gives them a 1.0% chance of having a particular type of cancer. Suppose we do a test: those testing positive now have a 35% chance of having cancer and those testing negative now have a 0.5% chance of having cancer. Based on the PPV of 35% one would be happy to refer the patient with a positive test to hospital to have a definitive diagnostic work-up. The 0.5% risk of cancer in those testing negative may not be sufficient to provide complete reassurance, but it is probably not great enough to justify expensive or invasive testing.
If on the other hand the same test is used in a screening context so that the risk of undiagnosed cancer in those tested is 0.04%, the PPV would be just 2.1% and the probability of disease in those testing negative would be 0.02%. In the screening setting, the risk of cancer in those testing positive may not be great enough to warrant further invasive testing. The value of the test for screening, depends critically on how those positive will be triaged. If the next step would be a CT scan then one might be willing to do 50 CT scans to diagnose one (asymptomatic) cancer, but if the next step if to perform surgery, one would probably not be willing to operate on 50 patients to diagnose one cancer.
The positive predictive value can be useful knowing what the options are for those who test positive, but one also needs to consider how the risk of disease changes as a result of the test and to realise that an acceptable value for the PPV depends critically on what comes next.
The views expressed are those of the author. Posting of the blog does not signify that the Cancer Prevention Group endorse those views or opinions.
Leave a Reply