By H. Gilbert Welch
Special to The Washington Post
Tuesday, February 19, 2008
In the effort to find a better way to identify which men will develop prostate cancer, the second most common cause of cancer death in men, the New England Journal of Medicine seems poised to announce a breakthrough. Later this month, the journal will publish a study showing how a combination of five gene variants -- so-called single nucleotide polymorphisms -- can help predict a man's risk of getting the disease. In the press release for the study, which has already been posted online, the authors wrote, "For the first time, this type of study has made it possible to develop a clinically viable gene test."
Genetic testing is a competitive area of research (the authors themselves are forming a company to market their test). But those of us who are potential consumers need to have a little healthy skepticism about the claim of clinical viability.
What does "viable" mean? It could mean that the test is able to be done. Or it could mean it's worth doing. If you are interested in the latter, here are three questions you'll want to ask:
What can the test tell you about your risk?
The study tested men with and without prostate cancer for 16 genetic variants. Although all 16 had been previously associated with prostate cancer by other researchers, this study showed that three variants had no relationship with the disease. This highlights an important caveat about genetic information: We now have the ability to read so many variants that some will appear to be related to a disease simply by chance. Before you believe a purported gene-disease association, you'll want to see it replicated in multiple studies.
The researchers went on to choose the five variants most strongly related to prostate cancer.
Even these five variants were, in the authors' own words, "only moderately associated with prostate cancer." Thus, they examined the cumulative effect of having multiple variants. The result was an apparently powerful association: Men who carry four (or all five) variants have 4.5 times the risk of developing prostate cancer compared with men who have none.
But it's rare for a man to have none of the five variants -- only 10 percent of men without prostate cancer had zero variants. The base case should reflect the typical man, and the typical man has two variants.
The base case used to calculate the change in risk matters a lot. Using zero gene variants as the base suggests that a man with any other test result is at an elevated risk. But, in fact, a man with two variants is at average risk.
And no one has their risk increased 4.5 fold. Instead, the biggest change is either a doubling of risk or a halving of risk.
Risk of what?
Getting prostate cancer isn't the same thing as dying from it. Prostate cancer is the poster child for an increasingly recognized side effect of early cancer detection: overdiagnosis, or the detection of cancers that will never matter to patients.
It turns out that there is a large reservoir of unsuspected prostate cancer. In some studies, more than half of men over age 60 have some pathological evidence of the disease. With the advent of PSA (prostate-specific antigen) screening, we are beginning to tap this reservoir and are now undoubtedly detecting and treating some men for "cancers" that would never have caused them problems during their life span. In short, many more men are being told they have prostate cancer and are being treated for a disease that would never kill them or even make them symptomatic.
The NEJM study shows that this genetic test won't fix the overdiagnosis problem.
In the study, some of the men with prostate cancer had very aggressive cancers; others had slow-growing cancers. But the genetic variants could not distinguish between the two. Nor could they predict which patients would develop prostate cancer at a young age or tell which patients already had cancers that had spread beyond the prostate. They could not predict the Gleason score (the pathologist's measure of how aggressive cells look under the microscope) or the PSA level (the biochemical measure of cancer aggressiveness). The variants did not even distinguish between those who had a family history of prostate cancer and those who did not (surprisingly, since one would expect this information to be contained in the genome).
In other words, while the test seemed to help predict the risk of getting prostate cancer, it didn't seem to help predict who would get a bad prostate cancer -- one that might kill you.
Does the risk information help?
It is perfectly plausible that a subsequent study will identify a set of genetic variants that will predict prostate cancer death. Imagine this was the study that did it.
Some men would be told their risk of prostate cancer death had doubled; others would be told their risk was cut in half. But the most common test results would be smaller changes in risk (for example, a risk that is 13 percent higher or, say, 23 percent lower than the risk of the average man). You would want to ground this information with the absolute risk of prostate cancer death -- which is roughly 3 percent over a lifetime.
So, you could learn that your risk of prostate cancer death over a lifetime was either 1.5 percent, 2.3 percent, 3 percent, 3.2 percent or 6.8 percent. You might also want to consider the competing risk of dying from something other than prostate cancer. That would still be more than 90 percent, no matter what your test result is.
Imagine you were a man in his 40s getting these results. What would you do differently? Would a lifetime risk of 6.8 percent be high enough to warrant a prophylactic prostatectomy? Certainly not for most of us.
Would you start hormonal therapy? Probably not, given the side effects of impotence, hot flashes, osteoporosis and heart disease. So that leaves you with looking for the disease early, through more regular PSA screening. The truth is we don't know whether PSA screening can help lower the death rate from prostate cancer.
If you believe it can, as many men and doctors do, wouldn't you continue to be tested regardless of whether your risk was 6.8 percent, 1.5 percent or anywhere in between? And if you are not a believer and are worried about the overdiagnosis and overtreatment initiated by PSA screening, would changes in your risk change your mind about screening?
You have to decide whether this test is viable or not. But you need to be ready for this new world of genetic testing. It's a world filled with tests that might show relatively small changes in risk -- a slightly higher risk of one cancer, a slightly lower risk of another, perhaps with other small changes in the risk of heart disease, diabetes, etc.
Exactly what we are predicting is not always clear. And what to do about what we find is even less so.
H. Gilbert Welch is a general internist at the Veterans Affairs Medical Center in White River Junction, Vt., and a professor at the Dartmouth Institute for Health Policy & Clinical Practice in Hanover, N.H. He is the author of "Should I Be Tested for Cancer?" (UC Press).