The New York Times (registration required) has an interesting article about the new test for EPO in urine samples, which is not a straightforward chemical detection and requires considerable interpretation by a skilled practitioner:
To distinguish recombinant EPO from the natural substance, the scientists used a technique that did not rely heavily on more straightforward chemical tests. To test for EPO, a preparation of urine is placed on the edge of a blotter, then subjected to the pull of an electrical field, leaving deposits in certain patterns that resemble a tiger's stripes. Some bands are more associated with recombinant EPO, some more with the natural substance. But there is considerable overlap as well. Scientists assess the patterns and the intensity of the bands. "There is a fair amount of cross-reactivity with the test, so there are certainly false positives," [Australian research scientist Nicolle] Packer ... said. "There is a lot of argument - and a lot of politics - about whether the test is good."
It should go without saying that if a positive test is going to be used to interrupt or terminate an athlete's career, invalidate past results, and ruin reputations, then the probability of false alarm (false positive rate) has to be very, very low. It doesn't sound to me like the EPO test meets that standard yet.
The story is inspired by the newest Lance Armstrong allegations, but it has broad applicability since EPO abuse for sports doping is thought to be very widespread. I found the article via Transition Game. You might also be interested in Nick's post, which is titled Dick Pound Must Go.