View From the Bushes:

I'm sure that most of you were hoping that I was done talking about cycling for the year, or at least that I was done talking about professional cycling and I was going to restrict myself to talking about the appallingly ugly rides I was attempting on my own. Well, unfortunately I couldn't let the most recent cycling story pass (in fact, I received an e-mail requesting that I talk about it). So the story as I reconstruct it goes something like this: whenever an athlete is tested they take two samples, an A and a B sample, so that if the A sample is positive they can test the B sample. Well it turns out that some B samples from the 1999 Tour were still hanging around, and that's where our story starts.

So France's doping lab, in an effort to improve the EPO test, unthaws some of this five-year old urine and starts running tests on it, and some of it comes up positive for EPO (using a test not available in 1999). At this point there's still no story because all the lab has is a number, but then the French newspaper L'Equipe comes along and shows a record with Armstrong's name on it which has a number that matches the number on one of the positive samples, and viola! Scandal erupts.

I work in the lab that did the testing for the 2002 Olympics, and while not a technician myself, I have some familiarity with what goes on and I have my doubts that you can exercise the same confidence in results you get off five-year old urine as the results you get off of "fresh" urine. And I'm not alone in this skepticism. Of course, since a) he's already retired, b) they don't have another sample to do the required confirmation and c) since only L'Equipe and not the lab itself is saying it's Armstrong, there's really nothing to be done except scream a lot.

My big picture take on this whole thing is that yes, there is some doping in cycling, but mostly by the "second tier" people who have historically performed well but never been able to reach the summit of the sport, people like Virenque, Dario Frigo, Rumsas, etc. I think the people really at the pinnacle, particularly people who have been there for quite awhile, have too much attention placed on them to ever consider doping. And when you look at the people I've listed, most of the time they've been caught by raids, not by tests. When tests do come back positive on some top-tier cyclist, it's always a new test or a test done in a really strange fashion, like the Armstrong test. So I guess the moral would be that the people doing the testing are being overzealous and using unproven tests and questionable methadology in an an attempt to stay ahead of the dopers and as a result, they end up seeing postives where there are none.

If only I biked as much as I talked about biking
Ross

Posted by direkobold at August 24, 2005 11:29 AM

Any testing regimen will produce false positives. This is a statistical reality that is glossed over by testing agencies and completely ignored by the media. I don't have numbers for EPO tests, but I do have some pretty good numbers for AIDS screening. (I suggest that all but math masochists stop reading here.)

Among men with no known risk factors (not an IV drug user, not homosexual) the incident of HIV infection is 0.01%. This is known as the base rate. If someone from this population who IS INFECTED with HIV is tested (one blood sample using the ELISA test confirmed by the Western Blot test), there is a 99.9% chance that the test will confirm they are positive. 99.9% is the test's sensitivity. If a man from this population that is NOT INFECTED is tested by the same protocol there is a 99.99% chance that the test will be negative. 99.99% is the test's specificity. With all those 9's it sounds like the test is foolproof. But let's say an employer test's a bunch of men with no known risk factors and fires everyone that tests positive (this was actually done by several large companies in Bombay). Imagine 10,000 are tested. Of these, 1 has HIV and will almost certainly test positive (99.9%). Of the other 9,999 men tested another one will also test postive because being right 99,99% of the time also means you are wrong once in every 10,000 tests. So bot men are fired, but only one actually had HIV. In statistics this is known as Bayes rule (usually stated a little differently). Basically it says that even with very sensitive, specific tests your certainty that a positive test is not a false positive is really low when you have a low base rate.

I doubt that the doping tests are anywhere near as sensitive and specific as the AIDS tests and they do do a huge amount of screeing compared to the number of positive test results that are produced. Ethically, without a backup sample, you cannot conclude anything. Even with duplicate tests it is well to be skeptical.

Posted by: sti at August 25, 2005 03:29 PM

I don't think the French will rest until Armstrong's name is dragged through the mud. I haven't canonized Armstrong like most of the media, but I still give him the benefit of the doubt over any French newspaper zealot.

Posted by: Ed at August 24, 2005 04:48 PM

It almost seems like we have someone in the family that works for a company that does blood tests that might have more information on the particular blood test they are talking about...now who is that again?

Posted by: Fred at August 25, 2005 08:15 PM