Tuesday, March 23, 2010

Imperfect Evidence and Information for the Public

Dr. Richard Ablin, the medical researcher who discovered the prostate specific antigen in 1970, wrote an op/ed in The New York Times a couple weeks ago in which he argues that we are wating a tremendous amount of money on PSA tests for men. There is a slight correlation between elevated PSA levels and prostate cancer, but some cases of the deadly form do not raise the PSA number. More prevalent is the false positive, the case in which the PSA number goes up for a reason other than cancer. Dr. Ablin writes,

"the test is hardly more effective than a coin toss. As I've been trying to make clear for many years now, P.S.A. testing can't detect prostate cancer and, more important, it can't distinguish between the two types of prostate cancer -- the one that will kill you and the one that won't."
My own doctor says that she attaches little importance to the PSA number and only considers it as evidence in context with other factors and tests (As my buddy Todd says, "They call it a digital exam, but it sure feels like analog to me").

But my doctor is a trained professional who thinks deeply about what one can and cannot infer from imperfect evidence, what about those who are not so trained? We give imperfect evidence to people all the time, sometimes for very important cases. Take, for example, pregnancy tests. What level of false negatives and false positives would be acceptable in such a test. You can try to explain it all you want in the instruction booklet, but you know that people who are worried and hiding the test, some who may have only a high school education, will not read it. How exact should the tests be to responsibly put them out for the public?