Drug Testing, Cycling, and Lance Armstrong

Update, October 11, 2012: You can read the full USADA decision, public statement, witness affidavits, and see other supporting documentation here. I haven't yet read any of it but press reports make the report sound pretty conclusive.

Are you wondering how it is that the United States Anti-Doping Agency (USADA) has decided to make a case against Lance Armstrong and the team leader, trainer, and team doctors despite all the years of denials and passed drug tests? (You can read a copy of the USADA charging letter here.) Even though the federal investigation was closed without charges?

One element to consider is that "passing" a drug test isn't the same thing as testing negative. As Kaiser Fung explains it in his excellent book Numbers Rule Your World:

Statistical testing shows that in steroid testing, a negative finding has far less value that a positive . . . for each doper caught red-handed (a true positive), one should expect about ten others to have escaped scot-free (false negatives). . . in particular, pay attention to these two numbers: the proportion of samples declared positive by drug-testing laboratories and the proportion of athletes believe to be using steroids. . .

 The Danish rider Bjarne Riis, who won the Tour de France in 1996, never tested positive either, yet he eventually admitted to using EPO, HGH, and cortisone. Marion Jones, the sprinter, denied drug use for years, until she eventually admitted it. The B sample [each sample is divided into two, with the second tested only if the first tests positive], Fung points out, acts as a protection against false positives. But the real issue is false negatives. As Fung says:

Statisticians say even if a test committed zero false positives, it would be far from "100 percent accurate" because of false-negative errors. Athletes only complain about false positives; the media wax on about false positives. We are missing the big story of steroid testing: false negatives. . . .

 The testers are timid because of the asymmetric costs from the two types of error. A false positive . . . will be rigorously litigated by the accused. An over-turned positive publicly humiliates the anti-doping authorities and diminishes the credibility of the testing program. By contrast, negative findings can only be proven false if athletes, like Riis, step forward to confess, so most false negatives never see the light of day.

Another element is the threshold point, that is, the dividing line between a positive and a negative screen. Where it is placed is up to the testers. Fung's example is the hematocrit level used during the 1990s: 50%. Normal is about 45%, but cycling used the higher number because some people, notably those from higher elevations, do have a higher level. "If the testers had used 60 percent, they could have reduced false positives, but without a doubt, more dopers would have evaded detection. Similarly, lowering the disqualifying hematocrit level would decrease false negatives at the expense of allowing more false positives." He concludes that testers have a structural reason for setting the bar high. "Because false positives make for lousy publicity, containing these errors is a top priority. However, this policy inevitably means some drug cheats are let loose."

So negative drug tests do not necessarily mean anything. The USADA charges go farther than simple drug use, implying a systematic plan by team management. The letter refers to eye-witnesses, but doesn't identify them. Yesterday there were reports that four of Armstrong's former teammates have withdrawn their names from consideration for the US Olympic team.